rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
postgresql/src/backend/access/transam/xact.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

6268 lines
178 KiB
C
Raw Normal View History

Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*-------------------------------------------------------------------------
*
Change my-function-name-- to my_function_name, and optimizer renames.
1999-02-14 00:22:53 +01:00
* xact.c
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
* top level transaction system support routines
*
Update the in-code documentation about the transaction system. Move it into a README file instead of being in xact.c's header comment. Alvaro Herrera.
2004-08-01 22:57:59 +02:00
* See src/backend/access/transam/README for more information.
*
Update copyright for 2023 Backpatch-through: 11
2023-01-02 21:00:37 +01:00
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
Add: * Portions Copyright (c) 1996-2000, PostgreSQL, Inc to all files copyright Regents of Berkeley. Man, that's a lot of files.
2000-01-26 06:58:53 +01:00
* Portions Copyright (c) 1994, Regents of the University of California
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*
* IDENTIFICATION
Remove cvs keywords from all files.
2010-09-20 22:08:53 +02:00
* src/backend/access/transam/xact.c
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*-------------------------------------------------------------------------
*/
-Wall'd That finishes access/*
1996-10-21 09:15:18 +02:00
Change #include's to use <> and "" as appropriate.
1999-07-16 01:04:24 +02:00
#include "postgres.h"
-Wall'd That finishes access/*
1996-10-21 09:15:18 +02:00
Adjust our timezone library to use pg_time_t (typedef'd as int64) in place of time_t, as per prior discussion. The behavior does not change on machines without a 64-bit-int type, but on machines with one, which is most, we are rid of the bizarre boundary behavior at the edges of the 32-bit-time_t range (1901 and 2038). The system will now treat times over the full supported timestamp range as being in your local time zone. It may seem a little bizarre to consider that times in 4000 BC are PST or EST, but this is surely at least as reasonable as propagating Gregorian calendar rules back that far. I did not modify the format of the zic timezone database files, which means that for the moment the system will not know about daylight-savings periods outside the range 1901-2038. Given the way the files are set up, it's not a simple decision like 'widen to 64 bits'; we have to actually think about the range of years that need to be supported. We should probably inquire what the plans of the upstream zic people are before making any decisions of our own.
2004-06-03 04:08:07 +02:00
#include <time.h>
Add missing #include.
2001-11-01 07:17:01 +01:00
#include <unistd.h>
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
#include "access/commit_ts.h"
Implement sharable row-level locks, and use them for foreign key references to eliminate unnecessary deadlocks. This commit adds SELECT ... FOR SHARE paralleling SELECT ... FOR UPDATE. The implementation uses a new SLRU data structure (managed much like pg_subtrans) to represent multiple- transaction-ID sets. When more than one transaction is holding a shared lock on a particular row, we create a MultiXactId representing that set of transactions and store its ID in the row's XMAX. This scheme allows an effectively unlimited number of row locks, just as we did before, while not costing any extra overhead except when a shared lock actually has to be shared. Still TODO: use the regular lock manager to control the grant order when multiple backends are waiting for a row lock. Alvaro Herrera and Tom Lane.
2005-04-28 23:47:18 +02:00
#include "access/multixact.h"
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
#include "access/parallel.h"
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
#include "access/subtrans.h"
Allow include files to compile own their own. Strip unused include files out unused include files, and add needed includes to C files. The next step is to remove unused include files in C files.
2006-07-13 18:49:20 +02:00
#include "access/transam.h"
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
#include "access/twophase.h"
Implement COMMIT_SIBLINGS parameter to allow pre-commit delay to occur only if at least N other backends currently have open transactions. This is not a great deal of intelligence about whether a delay might be profitable ... but it beats no intelligence at all. Note that the default COMMIT_DELAY is still zero --- this new code does nothing unless that setting is changed. Also, mark ENABLEFSYNC as a system-wide setting. It's no longer safe to allow that to be set per-backend, since we may be relying on some other backend's fsync to have synced the WAL log.
2001-02-26 01:50:08 +01:00
#include "access/xact.h"
Move the backup-block logic from XLogInsert to a new file, xloginsert.c. xlog.c is huge, this makes it a little bit smaller, which is nice. Functions related to putting together the WAL record are in xloginsert.c, and the lower level stuff for managing WAL buffers and such are in xlog.c. Also move the definition of XLogRecord to a separate header file. This causes churn in the #includes of all the files that write WAL records, and redo routines, but it avoids pulling in xlog.h into most places. Reviewed by Michael Paquier, Alvaro Herrera, Andres Freund and Amit Kapila.
2014-11-06 12:52:08 +01:00
#include "access/xlog.h"
#include "access/xloginsert.h"
Split xlog.c into xlog.c and xlogrecovery.c. This moves the functions related to performing WAL recovery into the new xlogrecovery.c source file, leaving xlog.c responsible for maintaining the WAL buffers, coordinating the startup and switch from recovery to normal operations, and other miscellaneous stuff that have always been in xlog.c. Reviewed-by: Andres Freund, Kyotaro Horiguchi, Robert Haas Discussion: https://www.postgresql.org/message-id/a31f27b4-a31d-f976-6217-2b03be646ffa%40iki.fi
2022-02-16 08:30:38 +01:00
#include "access/xlogrecovery.h"
Remove 576 references of include files that were not needed.
2006-07-14 16:52:27 +02:00
#include "access/xlogutils.h"
Fix possible crash during FATAL exit from reindexing. index.c supposed that it could just use a PG_TRY block to clean up the state associated with an active REINDEX operation. However, that code doesn't run if we do a FATAL exit --- for example, due to a SIGTERM shutdown signal --- while the REINDEX is happening. And that state does get consulted during catalog accesses, which makes it problematic if we do any catalog accesses during shutdown --- for example, to clean up any temp tables created in the session. If this combination of circumstances occurred, we could find ourselves trying to access already-freed memory. In debug builds that'd fairly reliably cause an assertion failure. In production we might often get away with it, but with some bad luck it could cause a core dump. Another possible bad outcome is an erroneous conclusion that an index-to-be-accessed is being reindexed; but it looks like that would be unlikely to have any consequences worse than failing to drop temp tables right away. (They'd still get dropped by the next session that uses that temp schema.) To fix, get rid of the use of PG_TRY here, and instead hook into the transaction abort mechanisms to clean up reindex state. Per bug #16378 from Alexander Lakhin. This has been wrong for a very long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/16378-7a70ca41b3ec2009@postgresql.org
2020-04-21 21:58:42 +02:00
#include "catalog/index.h"
Support temporary setting of search path during CREATE SCHEMA; this allows the example in the CREATE SCHEMA ref page to actually work now. Also, clean up when the transaction that initially creates a temp-table namespace is later aborted. Simplify internal representation of search path by folding special cases into the main list.
2002-05-17 22:53:33 +02:00
#include "catalog/namespace.h"
Relax transactional restrictions on ALTER TYPE ... ADD VALUE (redux). Originally committed as 15bc038f (plus some follow-ups), this was reverted in 28e07270 due to a problem discovered in parallel workers. This new version corrects that problem by sending the list of uncommitted enum values to parallel workers. Here follows the original commit message describing the change: To prevent possibly breaking indexes on enum columns, we must keep uncommitted enum values from getting stored in tables, unless we can be sure that any such column is new in the current transaction. Formerly, we enforced this by disallowing ALTER TYPE ... ADD VALUE from being executed at all in a transaction block, unless the target enum type had been created in the current transaction. This patch removes that restriction, and instead insists that an uncommitted enum value can't be referenced unless it belongs to an enum type created in the same transaction as the value. Per discussion, this should be a bit less onerous. It does require each function that could possibly return a new enum value to SQL operations to check this restriction, but there aren't so many of those that this seems unmaintainable. Author: Andrew Dunstan and Tom Lane, with parallel query fix by Thomas Munro Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAEepm%3D0Ei7g6PaNTbcmAh9tCRahQrk%3Dr5ZWLD-jr7hXweYX3yg%40mail.gmail.com Discussion: https://postgr.es/m/4075.1459088427%40sss.pgh.pa.us
2018-10-09 01:51:01 +02:00
#include "catalog/pg_enum.h"
Rethink the way FSM truncation works. Instead of WAL-logging FSM truncations in FSM code, call FreeSpaceMapTruncateRel from smgr_redo. To make that cleaner from modularity point of view, move the WAL-logging one level up to RelationTruncate, and move RelationTruncate and all the related WAL-logging to new src/backend/catalog/storage.c file. Introduce new RelationCreateStorage and RelationDropStorage functions that are used instead of calling smgrcreate/smgrscheduleunlink directly. Move the pending rel deletion stuff from smgrcreate/smgrscheduleunlink to the new functions. This leaves smgr.c as a thin wrapper around md.c; all the transactional stuff is now in storage.c. This will make it easier to add new forks with similar truncation logic, like the visibility map.
2008-11-19 11:34:52 +01:00
#include "catalog/storage.h"
Change #include's to use <> and "" as appropriate.
1999-07-16 01:04:24 +02:00
#include "commands/async.h"
Add code to handle [ON COMMIT { PRESERVE ROWS | DELETE ROWS | DROP }] for temp tables. Gavin Sherry
2002-11-10 00:56:39 +01:00
#include "commands/tablecmds.h"
This is part #1 for of the DEFERRED CONSTRAINT TRIGGER support. Implements the CREATE CONSTRAINT TRIGGER and SET CONSTRAINTS commands. TODO: Generic builtin trigger procedures Automatic execution of appropriate CREATE CONSTRAINT... at CREATE TABLE Support of new trigger type in pg_dump Swapping of huge # of events to disk Jan
1999-09-29 18:06:40 +02:00
#include "commands/trigger.h"
Replace random(), pg_erand48(), etc with a better PRNG API and algorithm. Standardize on xoroshiro128** as our basic PRNG algorithm, eliminating a bunch of platform dependencies as well as fundamentally-obsolete PRNG code. In addition, this API replacement will ease replacing the algorithm again in future, should that become necessary. xoroshiro128** is a few percent slower than the drand48 family, but it can produce full-width 64-bit random values not only 48-bit, and it should be much more trustworthy. It's likely to be noticeably faster than the platform's random(), depending on which platform you are thinking about; and we can have non-global state vectors easily, unlike with random(). It is not cryptographically strong, but neither are the functions it replaces. Fabien Coelho, reviewed by Dean Rasheed, Aleksander Alekseev, and myself Discussion: https://postgr.es/m/alpine.DEB.2.22.394.2105241211230.165418@pseudo
2021-11-29 03:32:36 +01:00
#include "common/pg_prng.h"
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
#include "executor/spi.h"
Change #include's to use <> and "" as appropriate.
1999-07-16 01:04:24 +02:00
#include "libpq/be-fsstubs.h"
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
#include "libpq/pqsignal.h"
Reset current user id to session user id during transaction abort
2000-09-27 12:41:55 +02:00
#include "miscadmin.h"
Make the order of the header file includes consistent in backend modules. Similar to commits 7e735035f2 and dddf4cdc33, this commit makes the order of header file inclusion consistent for backend modules. In the passing, removed a couple of duplicate inclusions. Author: Vignesh C Reviewed-by: Kuntal Ghosh and Amit Kapila Discussion: https://postgr.es/m/CALDaNm2Sznv8RR6Ex-iJO6xAdsxgWhCoETkaYX=+9DW3q0QCfA@mail.gmail.com
2019-11-12 04:00:16 +01:00
#include "pg_trace.h"
Remove 576 references of include files that were not needed.
2006-07-14 16:52:27 +02:00
#include "pgstat.h"
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
#include "replication/logical.h"
Logical replication - Add PUBLICATION catalogs and DDL - Add SUBSCRIPTION catalog and DDL - Define logical replication protocol and output plugin - Add logical replication workers From: Petr Jelinek <petr@2ndquadrant.com> Reviewed-by: Steve Singer <steve@ssinger.info> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Erik Rijkers <er@xs4all.nl> Reviewed-by: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>
2017-01-19 18:00:00 +01:00
#include "replication/logicallauncher.h"
Wake up a subscription's replication worker processes after DDL. Waken related worker processes immediately at commit of a transaction that has performed ALTER SUBSCRIPTION (including the RENAME and OWNER variants). This reduces the response time for such operations. In the real world that might not be worth much, but it shaves several seconds off the runtime for the subscription test suite. In the case of PREPARE, we just throw away this notification state; it doesn't seem worth the work to preserve it. The workers will still react after the eventual COMMIT PREPARED, but not as quickly. Nathan Bossart Discussion: https://postgr.es/m/20221122004119.GA132961@nathanxps13
2023-01-06 22:08:20 +01:00
#include "replication/logicalworker.h"
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
#include "replication/origin.h"
Reset properly snapshot export state during transaction abort During a replication slot creation, an ERROR generated in the same transaction as the one creating a to-be-exported snapshot would have left the backend in an inconsistent state, as the associated static export snapshot state was not being reset on transaction abort, but only on the follow-up command received by the WAL sender that created this snapshot on replication slot creation. This would trigger inconsistency failures if this session tried to export again a snapshot, like during the creation of a replication slot. Note that a snapshot export cannot happen in a transaction block, so there is no need to worry resetting this state for subtransaction aborts. Also, this inconsistent state would very unlikely show up to users. For example, one case where this could happen is an out-of-memory error when building the initial snapshot to-be-exported. Dilip found this problem while poking at a different patch, that caused an error in this code path for reasons unrelated to HEAD. Author: Dilip Kumar Reviewed-by: Michael Paquier, Zhihong Yu Discussion: https://postgr.es/m/CAFiTN-s0zA1Kj0ozGHwkYkHwa5U0zUE94RSc_g81WrpcETB5=w@mail.gmail.com Backpatch-through: 9.6
2021-10-18 04:55:42 +02:00
#include "replication/snapbuild.h"
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
#include "replication/syncrep.h"
#include "replication/walsender.h"
Support condition variables. Condition variables provide a flexible way to sleep until a cooperating process causes an arbitrary condition to become true. In simple cases, this can be accomplished with a WaitLatch/ResetLatch loop; the cooperating process can call SetLatch after performing work that might cause the condition to be satisfied, and the waiting process can recheck the condition each time. However, if the process performing the work doesn't have an easy way to identify which processes might be waiting, this doesn't work, because it can't identify which latches to set. Condition variables solve that problem by internally maintaining a list of waiters; a process that may have caused some waiter's condition to be satisfied must "signal" or "broadcast" on the condition variable. Robert Haas and Thomas Munro
2016-11-22 20:26:40 +01:00
#include "storage/condition_variable.h"
Split resowner.h This lets files that are mere users of ResourceOwner not automatically include the headers for stuff that is managed by the resowner mechanism.
2012-08-29 00:02:07 +02:00
#include "storage/fd.h"
Remove 576 references of include files that were not needed.
2006-07-14 16:52:27 +02:00
#include "storage/lmgr.h"
Refactor the fsync queue for wider use. Previously, md.c and checkpointer.c were tightly integrated so that fsync calls could be handed off and processed in the background. Introduce a system of callbacks and file tags, so that other modules can hand off fsync work in the same way. For now only md.c uses the new interface, but other users are being proposed. Since there may be use cases that are not strictly SMGR implementations, use a new function table for sync handlers rather than extending the traditional SMGR one. Instead of using a bitmapset of segment numbers for each RelFileNode in the checkpointer's hash table, make the segment number part of the key. This requires sending explicit "forget" requests for every segment individually when relations are dropped, but suits the file layout schemes of proposed future users better (ie sparse or high segment numbers). Author: Shawn Debnath and Thomas Munro Reviewed-by: Thomas Munro, Andres Freund Discussion: https://postgr.es/m/CAEepm=2gTANm=e3ARnJT=n0h8hf88wqmaZxk0JYkxw+b21fNrw@mail.gmail.com
2019-04-04 10:56:03 +02:00
#include "storage/md.h"
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
#include "storage/predicate.h"
Tighten up includes in sinvaladt.h, twophase.h, proc.h Remove proc.h from sinvaladt.h and twophase.h; also replace xlog.h in proc.h with xlogdefs.h.
2012-06-25 23:45:15 +02:00
#include "storage/proc.h"
Split the shared-memory array of PGPROC pointers out of the sinval communication structure, and make it its own module with its own lock. This should reduce contention at least a little, and it definitely makes the code seem cleaner. Per my recent proposal.
2005-05-19 23:35:48 +02:00
#include "storage/procarray.h"
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
#include "storage/sinvaladt.h"
Make DROP TABLE rollback-able: postpone physical file delete until commit. (WAL logging for this is not done yet, however.) Clean up a number of really crufty things that are no longer needed now that DROP behaves nicely. Make temp table mapper do the right things when drop or rename affecting a temp table is rolled back. Also, remove "relation modified while in use" error check, in favor of locking tables at first reference and holding that lock throughout the statement.
2000-11-08 23:10:03 +01:00
#include "storage/smgr.h"
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
#include "utils/builtins.h"
Split resowner.h This lets files that are mere users of ResourceOwner not automatically include the headers for stuff that is managed by the resowner mechanism.
2012-08-29 00:02:07 +02:00
#include "utils/catcache.h"
Combine cmin and cmax fields of HeapTupleHeaders into a single field, by keeping private state in each backend that has inserted and deleted the same tuple during its current top-level transaction. This is sufficient since there is no need to be able to determine the cmin/cmax from any other transaction. This gets us back down to 23-byte headers, removing a penalty paid in 8.0 to support subtransactions. Patch by Heikki Linnakangas, with minor revisions by moi, following a design hashed out awhile back on the pghackers list.
2007-02-09 04:35:35 +01:00
#include "utils/combocid.h"
Clean up the #include mess a little. walsender.h should depend on xlog.h, not vice versa. (Actually, the inclusion was circular until a couple hours ago, which was even sillier; but Bruce broke it in the expedient rather than logically correct direction.) Because of that poor decision, plus blind application of pgrminclude, we had a situation where half the system was depending on xlog.h to include such unrelated stuff as array.h and guc.h. Clean up the header inclusion, and manually revert a lot of what pgrminclude had done so things build again. This episode reinforces my feeling that pgrminclude should not be run without adult supervision. Inclusion changes in header files in particular need to be reviewed with great care. More generally, it'd be good if we had a clearer notion of module layering to dictate which headers can sanely include which others ... but that's a big task for another day.
2011-09-04 07:13:16 +02:00
#include "utils/guc.h"
Final cleanup.
1999-07-16 07:00:38 +02:00
#include "utils/inval.h"
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
#include "utils/memutils.h"
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
#include "utils/relmapper.h"
Rename snapmgmt.c/h to snapmgr.c/h, for consistency with other files. Per complaint from Tom Lane.
2008-03-26 19:48:59 +01:00
#include "utils/snapmgr.h"
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
#include "utils/timeout.h"
Move Timestamp/Interval typedefs and basic macros into datatype/timestamp.h. As per my recent proposal, this refactors things so that these typedefs and macros are available in a header that can be included in frontend-ish code. I also changed various headers that were undesirably including utils/timestamp.h to include datatype/timestamp.h instead. Unsurprisingly, this showed that half the system was getting utils/timestamp.h by way of xlog.h. No actual code changes here, just header refactoring.
2011-09-09 19:23:41 +02:00
#include "utils/timestamp.h"
Reimplement temp tables using schemas. The temp table map is history; temp table entries in pg_class have the names the user would expect.
2002-03-31 08:26:32 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* User-tweakable parameters
*/
int DefaultXactIsoLevel = XACT_READ_COMMITTED;
Clean up some inconsistencies with GUC declarations This is similar to 7d25958, and this commit takes care of all the remaining inconsistencies between the initial value used in the C variable associated to a GUC and its default value stored in the GUC tables (as of pg_settings.boot_val). Some of the initial values of the GUCs updated rely on a compile-time default. These are refactored so as the GUC table and its C declaration use the same values. This makes everything consistent with other places, backend_flush_after, bgwriter_flush_after, port, checkpoint_flush_after doing so already, for example. Extracted from a larger patch by Peter Smith. The spots updated in the modules are from me. Author: Peter Smith, Michael Paquier Reviewed-by: Nathan Bossart, Tom Lane, Justin Pryzby Discussion: https://postgr.es/m/CAHut+PtHE0XSfjjRQ6D4v7+dqzCw=d+1a64ujra4EX8aoc_Z+w@mail.gmail.com
2022-10-31 04:44:48 +01:00
int XactIsoLevel = XACT_READ_COMMITTED;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
bool DefaultXactReadOnly = false;
bool XactReadOnly;
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
bool DefaultXactDeferrable = false;
bool XactDeferrable;
Merge synchronous_replication setting into synchronous_commit. This means one less thing to configure when setting up synchronous replication, and also avoids some ambiguity around what the behavior should be when the settings of these variables conflict. Fujii Masao, with additional hacking by me.
2011-04-04 22:13:01 +02:00
int synchronous_commit = SYNCHRONOUS_COMMIT_ON;
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
Implement streaming mode in ReorderBuffer. Instead of serializing the transaction to disk after reaching the logical_decoding_work_mem limit in memory, we consume the changes we have in memory and invoke stream API methods added by commit 45fdc9738b. However, sometimes if we have incomplete toast or speculative insert we spill to the disk because we can't generate the complete tuple and stream. And, as soon as we get the complete tuple we stream the transaction including the serialized changes. We can do this incremental processing thanks to having assignments (associating subxact with toplevel xacts) in WAL right away, and thanks to logging the invalidation messages at each command end. These features are added by commits 0bead9af48 and c55040ccd0 respectively. Now that we can stream in-progress transactions, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We handle such failures by returning ERRCODE_TRANSACTION_ROLLBACK sqlerrcode from system table scan APIs to the backend or WALSender decoding a specific uncommitted transaction. The decoding logic on the receipt of such a sqlerrcode aborts the decoding of the current transaction and continue with the decoding of other transactions. We have ReorderBufferTXN pointer in each ReorderBufferChange by which we know which xact it belongs to. The output plugin can use this to decide which changes to discard in case of stream_abort_cb (e.g. when a subxact gets discarded). We also provide a new option via SQL APIs to fetch the changes being streamed. Author: Dilip Kumar, Tomas Vondra, Amit Kapila, Nikhil Sontakke Reviewed-by: Amit Kapila, Kuntal Ghosh, Ajin Cherian Tested-by: Neha Sharma, Mahendra Singh Thalor and Ajin Cherian Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2020-08-08 04:04:39 +02:00
/*
* CheckXidAlive is a xid value pointing to a possibly ongoing (sub)
* transaction. Currently, it is used in logical decoding. It's possible
* that such transactions can get aborted while the decoding is ongoing in
* which case we skip decoding that particular transaction. To ensure that we
* check whether the CheckXidAlive is aborted after fetching the tuple from
* system tables. We also ensure that during logical decoding we never
* directly access the tableam or heap APIs because we are checking for the
* concurrent aborts only in systable_* APIs.
*/
TransactionId CheckXidAlive = InvalidTransactionId;
bool bsysscan = false;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* When running as a parallel worker, we place only a single
* TransactionStateData on the parallel worker's state stack, and the XID
* reflected there will be that of the *innermost* currently-active
Fix misc typos. Oskari Saarenmaa. Backpatch to stable branches where applicable.
2015-09-05 10:35:49 +02:00
* subtransaction in the backend that initiated parallelism. However,
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* GetTopTransactionId() and TransactionIdIsCurrentTransactionId()
* need to return the same answers in the parallel worker as they would have
* in the user backend, so we need some additional bookkeeping.
*
Fix assorted inconsistencies. There were a number of issues in the recent commits which include typos, code and comments mismatch, leftover function declarations. Fix them. Reported-by: Alexander Lakhin Author: Alexander Lakhin, Amit Kapila and Amit Langote Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/ef0c0232-0c1d-3a35-63d4-0ebd06e31387@gmail.com
2019-06-08 04:46:38 +02:00
* XactTopFullTransactionId stores the XID of our toplevel transaction, which
Fix inconsistencies and typos in the tree, take 11 This fixes various typos in docs and comments, and removes some orphaned definitions. Author: Alexander Lakhin Discussion: https://postgr.es/m/5da8e325-c665-da95-21e0-c8a99ea61fbf@gmail.com
2019-08-19 09:21:39 +02:00
* will be the same as TopTransactionStateData.fullTransactionId in an
* ordinary backend; but in a parallel backend, which does not have the entire
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* transaction state, it will instead be copied from the backend that started
* the parallel operation.
*
* nParallelCurrentXids will be 0 and ParallelCurrentXids NULL in an ordinary
* backend, but in a parallel backend, nParallelCurrentXids will contain the
* number of XIDs that need to be considered current, and ParallelCurrentXids
* will contain the XIDs themselves. This includes all XIDs that were current
* or sub-committed in the parent at the time the parallel operation began.
* The XIDs are stored sorted in numerical order (not logical order) to make
* lookups as fast as possible.
*/
Add 'static' to file-local variables missing it. Noticed when comparing the set of exported symbols without / with -fvisibility=hidden after adding PGDLLIMPORT to intentionally exported symbols. Discussion: https://postgr.es/m/20220512164513.vaheofqp2q24l65r@alap3.anarazel.de
2022-05-12 18:19:04 +02:00
static FullTransactionId XactTopFullTransactionId = {InvalidTransactionId};
static int nParallelCurrentXids = 0;
static TransactionId *ParallelCurrentXids;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
Fix PREPARE TRANSACTION to reject the case where the transaction has dropped a temporary table; we can't support that because there's no way to clean up the source backend's internal state if the eventual COMMIT PREPARED is done by another backend. This was checked correctly in 8.1 but I broke it in 8.2 :-(. Patch by Heikki Linnakangas, original trouble report by John Smith.
2008-03-04 20:54:06 +01:00
/*
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
* Miscellaneous flag bits to record events which occur on the top level
* transaction. These flags are only persisted in MyXactFlags and are intended
* so we remember to do certain things later on in the transaction. This is
* globally accessible, so can be set from anywhere in the code that requires
* recording flags.
Fix PREPARE TRANSACTION to reject the case where the transaction has dropped a temporary table; we can't support that because there's no way to clean up the source backend's internal state if the eventual COMMIT PREPARED is done by another backend. This was checked correctly in 8.1 but I broke it in 8.2 :-(. Patch by Heikki Linnakangas, original trouble report by John Smith.
2008-03-04 20:54:06 +01:00
*/
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
int MyXactFlags;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
* transaction states - transaction state from server perspective
*/
typedef enum TransState
{
Redefine IsTransactionState() to only return true for TRANS_INPROGRESS state, which is the only state in which it's safe to initiate database queries. It turns out that all but two of the callers thought that's what it meant; and the other two were using it as a proxy for "will GetTopTransactionId() return a nonzero XID"? Since it was in fact an unreliable guide to that, make those two just invoke GetTopTransactionId() always, then deal with a zero result if they get one.
2007-06-07 23:45:59 +02:00
TRANS_DEFAULT, /* idle */
TRANS_START, /* transaction starting */
TRANS_INPROGRESS, /* inside a valid transaction */
TRANS_COMMIT, /* commit in progress */
TRANS_ABORT, /* abort in progress */
TRANS_PREPARE /* prepare in progress */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
} TransState;
/*
* transaction block states - transaction state of client queries
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* Note: the subtransaction states are used only for non-topmost
* transactions; the others appear only in the topmost transaction.
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
*/
typedef enum TBlockState
{
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/* not-in-transaction-block states */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TBLOCK_DEFAULT, /* idle */
TBLOCK_STARTED, /* running single-query transaction */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/* transaction block states */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TBLOCK_BEGIN, /* starting transaction block */
TBLOCK_INPROGRESS, /* live transaction */
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
TBLOCK_IMPLICIT_INPROGRESS, /* live transaction after implicit BEGIN */
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
TBLOCK_PARALLEL_INPROGRESS, /* live transaction inside parallel worker */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TBLOCK_END, /* COMMIT received */
TBLOCK_ABORT, /* failed xact, awaiting ROLLBACK */
TBLOCK_ABORT_END, /* failed xact, ROLLBACK received */
TBLOCK_ABORT_PENDING, /* live xact, ROLLBACK received */
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
TBLOCK_PREPARE, /* live xact, PREPARE received */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/* subtransaction states */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TBLOCK_SUBBEGIN, /* starting a subtransaction */
TBLOCK_SUBINPROGRESS, /* live subtransaction */
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
TBLOCK_SUBRELEASE, /* RELEASE received */
TBLOCK_SUBCOMMIT, /* COMMIT received while TBLOCK_SUBINPROGRESS */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TBLOCK_SUBABORT, /* failed subxact, awaiting ROLLBACK */
TBLOCK_SUBABORT_END, /* failed subxact, ROLLBACK received */
TBLOCK_SUBABORT_PENDING, /* live subxact, ROLLBACK received */
TBLOCK_SUBRESTART, /* live subxact, ROLLBACK TO received */
TBLOCK_SUBABORT_RESTART /* failed subxact, ROLLBACK TO received */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
} TBlockState;
/*
* transaction state structure
*/
typedef struct TransactionStateData
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
FullTransactionId fullTransactionId; /* my FullTransactionId */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
SubTransactionId subTransactionId; /* my subxact ID */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
char *name; /* savepoint name, if any */
int savepointLevel; /* savepoint level */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
TransState state; /* low-level state */
TBlockState blockState; /* high-level state */
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
int nestingLevel; /* transaction nesting depth */
int gucNestLevel; /* GUC context nesting depth */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
MemoryContext curTransactionContext; /* my xact-lifetime context */
ResourceOwner curTransactionOwner; /* my query resources */
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
TransactionId *childXids; /* subcommitted child XIDs, in XID order */
int nChildXids; /* # of subcommitted child XIDs */
int maxChildXids; /* allocated size of childXids[] */
Make standard maintenance operations (including VACUUM, ANALYZE, REINDEX, and CLUSTER) execute as the table owner rather than the calling user, using the same privilege-switching mechanism already used for SECURITY DEFINER functions. The purpose of this change is to ensure that user-defined functions used in index definitions cannot acquire the privileges of a superuser account that is performing routine maintenance. While a function used in an index is supposed to be IMMUTABLE and thus not able to do anything very interesting, there are several easy ways around that restriction; and even if we could plug them all, there would remain a risk of reading sensitive information and broadcasting it through a covert channel such as CPU usage. To prevent bypassing this security measure, execution of SET SESSION AUTHORIZATION and SET ROLE is now forbidden within a SECURITY DEFINER context. Thanks to Itagaki Takahiro for reporting this vulnerability. Security: CVE-2007-6600
2008-01-03 22:23:15 +01:00
Oid prevUser; /* previous CurrentUserId setting */
Prevent indirect security attacks via changing session-local state within an allegedly immutable index function. It was previously recognized that we had to prevent such a function from executing SET/RESET ROLE/SESSION AUTHORIZATION, or it could trivially obtain the privileges of the session user. However, since there is in general no privilege checking for changes of session-local state, it is also possible for such a function to change settings in a way that might subvert later operations in the same session. Examples include changing search_path to cause an unexpected function to be called, or replacing an existing prepared statement with another one that will execute a function of the attacker's choosing. The present patch secures VACUUM, ANALYZE, and CREATE INDEX/REINDEX against these threats, which are the same places previously deemed to need protection against the SET ROLE issue. GUC changes are still allowed, since there are many useful cases for that, but we prevent security problems by forcing a rollback of any GUC change after completing the operation. Other cases are handled by throwing an error if any change is attempted; these include temp table creation, closing a cursor, and creating or deleting a prepared statement. (In 7.4, the infrastructure to roll back GUC changes doesn't exist, so we settle for rejecting changes of "search_path" in these contexts.) Original report and patch by Gurjeet Singh, additional analysis by Tom Lane. Security: CVE-2009-4136
2009-12-09 22:57:51 +01:00
int prevSecContext; /* previous SecurityRestrictionContext */
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
bool prevXactReadOnly; /* entry-time xact r/o state */
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
bool startedInRecovery; /* did we start in recovery? */
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
bool didLogXid; /* has xid been included in WAL record? */
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
int parallelModeLevel; /* Enter/ExitParallelMode counter */
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
bool chain; /* start a new block after this one */
Replace XLOG_INCLUDE_XID flag with a more localized flag. Commit 0bead9af484c introduced XLOG_INCLUDE_XID flag to indicate that the WAL record contains subXID-to-topXID association. It uses that flag later to mark in CurrentTransactionState that top-xid is logged so that we should not try to log it again with the next WAL record in the current subtransaction. However, we can use a localized variable to pass that information. In passing, change the related function and variable names to make them consistent with what the code is actually doing. Author: Dilip Kumar Reviewed-by: Alvaro Herrera, Amit Kapila Discussion: https://postgr.es/m/E1mSoYz-0007Fh-D9@gemulon.postgresql.org
2021-11-02 04:05:29 +01:00
bool topXidLogged; /* for a subxact: is top-level XID logged? */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
struct TransactionStateData *parent; /* back link to parent */
} TransactionStateData;
typedef TransactionStateData *TransactionState;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
/*
* Serialized representation used to transmit transaction state to parallel
Fix typo. Author: Masahiko Sawada
2019-03-28 22:03:58 +01:00
* workers through shared memory.
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
*/
typedef struct SerializedTransactionState
{
int xactIsoLevel;
bool xactDeferrable;
FullTransactionId topFullTransactionId;
FullTransactionId currentFullTransactionId;
CommandId currentCommandId;
int nParallelCurrentXids;
TransactionId parallelCurrentXids[FLEXIBLE_ARRAY_MEMBER];
} SerializedTransactionState;
/* The size of SerializedTransactionState, not including the final array. */
#define SerializedTransactionStateHeaderSize \
offsetof(SerializedTransactionState, parallelCurrentXids)
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
/*
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
* CurrentTransactionState always points to the current transaction state
* block. It will point to TopTransactionStateData when not in a
* transaction at all, or when in a top-level transaction.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
static TransactionStateData TopTransactionStateData = {
Use C99 designated initializers for some structs These are just a few particularly egregious cases that were hard to read and write, and error prone because of many similar adjacent types. Discussion: https://www.postgresql.org/message-id/flat/4c9f01be-9245-2148-b569-61a8562ef190%402ndquadrant.com
2018-08-29 08:36:30 +02:00
.state = TRANS_DEFAULT,
.blockState = TBLOCK_DEFAULT,
Replace XLOG_INCLUDE_XID flag with a more localized flag. Commit 0bead9af484c introduced XLOG_INCLUDE_XID flag to indicate that the WAL record contains subXID-to-topXID association. It uses that flag later to mark in CurrentTransactionState that top-xid is logged so that we should not try to log it again with the next WAL record in the current subtransaction. However, we can use a localized variable to pass that information. In passing, change the related function and variable names to make them consistent with what the code is actually doing. Author: Dilip Kumar Reviewed-by: Alvaro Herrera, Amit Kapila Discussion: https://postgr.es/m/E1mSoYz-0007Fh-D9@gemulon.postgresql.org
2021-11-02 04:05:29 +01:00
.topXidLogged = false,
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
};
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* unreportedXids holds XIDs of all subtransactions that have not yet been
Collection of typo fixes. Use "a" and "an" correctly, mostly in comments. Two error messages were also fixed (they were just elogs, so no translation work required). Two function comments in pg_proc.h were also fixed. Etsuro Fujita reported one of these, but I found a lot more with grep. Also fix a few other typos spotted while grepping for the a/an typos. For example, "consists out of ..." -> "consists of ...". Plus a "though"/ "through" mixup reported by Euler Taveira. Many of these typos were in old code, which would be nice to backpatch to make future backpatching easier. But much of the code was new, and I didn't feel like crafting separate patches for each branch. So no backpatching.
2015-05-20 15:18:11 +02:00
* reported in an XLOG_XACT_ASSIGNMENT record.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
static int nUnreportedXids;
static TransactionId unreportedXids[PGPROC_MAX_CACHED_SUBXIDS];
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
static TransactionState CurrentTransactionState = &TopTransactionStateData;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* The subtransaction ID and command ID assignment counters are global
* to a whole transaction, so we do not keep them in the state stack.
*/
static SubTransactionId currentSubTransactionId;
static CommandId currentCommandId;
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
static bool currentCommandIdUsed;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Remove redundant gettimeofday() calls to the extent practical without changing semantics too much. statement_timestamp is now set immediately upon receipt of a client command message, and the various places that used to do their own gettimeofday() calls to mark command startup are referenced to that instead. I have also made stats_command_string use that same value for pg_stat_activity.query_start for both the command itself and its eventual replacement by <IDLE> or <idle in transaction>. There was some debate about that, but no argument that seemed convincing enough to justify an extra gettimeofday() call.
2006-06-21 00:52:00 +02:00
* xactStartTimestamp is the value of transaction_timestamp().
* stmtStartTimestamp is the value of statement_timestamp().
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
* xactStopTimestamp is the time at which we log a commit / abort WAL record,
* or if that was skipped, the time of the first subsequent
* GetCurrentTransactionStopTimestamp() call.
*
Remove redundant gettimeofday() calls to the extent practical without changing semantics too much. statement_timestamp is now set immediately upon receipt of a client command message, and the various places that used to do their own gettimeofday() calls to mark command startup are referenced to that instead. I have also made stats_command_string use that same value for pg_stat_activity.query_start for both the command itself and its eventual replacement by <IDLE> or <idle in transaction>. There was some debate about that, but no argument that seemed convincing enough to justify an extra gettimeofday() call.
2006-06-21 00:52:00 +02:00
* These do not change as we enter and exit subtransactions, so we don't
* keep them inside the TransactionState stack.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Clean up the rather historically encumbered interface to now() and current time: provide a GetCurrentTimestamp() function that returns current time in the form of a TimestampTz, instead of separate time_t and microseconds fields. This is what all the callers really want anyway, and it eliminates low-level dependencies on AbsoluteTime, which is a deprecated datatype that will have to disappear eventually.
2005-06-30 00:51:57 +02:00
static TimestampTz xactStartTimestamp;
Add statement_timestamp(), clock_timestamp(), and transaction_timestamp() (just like now()). Also update statement_timeout() to mention it is statement arrival time that is measured. Catalog version updated.
2006-04-25 02:25:22 +02:00
static TimestampTz stmtStartTimestamp;
Implement rate-limiting logic on how often backends will attempt to send messages to the stats collector. This avoids the problem that enabling stats_row_level for autovacuum has a significant overhead for short read-only transactions, as noted by Arjen van der Meijden. We can avoid an extra gettimeofday call by piggybacking on the one done for WAL-logging xact commit or abort (although that doesn't help read-only transactions, since they don't WAL-log anything). In my proposal for this, I noted that we could change the WAL log entries for commit/abort to record full TimestampTz precision, instead of only time_t as at present. That's not done in this patch, but will be committed separately.
2007-04-30 05:23:49 +02:00
static TimestampTz xactStopTimestamp;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* GID to be used for preparing the current transaction. This is also
* global to a whole transaction, so we don't keep it in the state stack.
*/
static char *prepareGID;
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
/*
* Some commands want to force synchronous commit.
*/
static bool forceSyncCommit = false;
Log all statements from a sample of transactions This is useful to obtain a view of the different transaction types in an application, regardless of the durations of the statements each runs. Author: Adrien Nayrat Reviewed-by: Masahiko Sawada, Hayato Kuroda, Andres Freund
2019-04-03 23:43:59 +02:00
/* Flag for logging statements in a transaction. */
bool xact_is_sampled = false;
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/*
* Private context for transaction-abort work --- we reserve space for this
* at startup to ensure that AbortTransaction and AbortSubTransaction can work
* when we've run out of memory.
*/
static MemoryContext TransactionAbortContext = NULL;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
/*
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
* List of add-on start- and end-of-xact callbacks
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
*/
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
typedef struct XactCallbackItem
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
{
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
struct XactCallbackItem *next;
XactCallback callback;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
void *arg;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
} XactCallbackItem;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
static XactCallbackItem *Xact_callbacks = NULL;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* List of add-on start- and end-of-subxact callbacks
*/
typedef struct SubXactCallbackItem
{
struct SubXactCallbackItem *next;
SubXactCallback callback;
void *arg;
} SubXactCallbackItem;
static SubXactCallbackItem *SubXact_callbacks = NULL;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* local function prototypes */
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
static void AssignTransactionId(TransactionState s);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void AbortTransaction(void);
static void AtAbort_Memory(void);
static void AtCleanup_Memory(void);
PortalRun must guard against the possibility that the portal it's running contains VACUUM or a similar command that will internally start and commit transactions. In such a case, the original caller values of CurrentMemoryContext and CurrentResourceOwner will point to objects that will be destroyed by the internal commit. We must restore these pointers to point to the newly-manufactured transaction context and resource owner, rather than possibly pointing to deleted memory. Also tweak xact.c so that AbortTransaction and AbortSubTransaction forcibly restore a sane value for CurrentResourceOwner, much as they have always done for CurrentMemoryContext. I'm not certain this is necessary but I'm feeling paranoid today. Responds to Sean Chittenden's bug report of 4-Oct.
2004-10-04 23:52:15 +02:00
static void AtAbort_ResourceOwner(void);
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
static void AtCCI_LocalCache(void);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void AtCommit_Memory(void);
static void AtStart_Cache(void);
static void AtStart_Memory(void);
static void AtStart_ResourceOwner(void);
static void CallXactCallbacks(XactEvent event);
static void CallSubXactCallbacks(SubXactEvent event,
SubTransactionId mySubid,
SubTransactionId parentSubid);
static void CleanupTransaction(void);
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
static void CheckTransactionBlock(bool isTopLevel, bool throwError,
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
const char *stmtType);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void CommitTransaction(void);
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
static TransactionId RecordTransactionAbort(bool isSubXact);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void StartTransaction(void);
static void StartSubTransaction(void);
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
static void CommitSubTransaction(void);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void AbortSubTransaction(void);
static void CleanupSubTransaction(void);
static void PushTransaction(void);
static void PopTransaction(void);
static void AtSubAbort_Memory(void);
static void AtSubCleanup_Memory(void);
PortalRun must guard against the possibility that the portal it's running contains VACUUM or a similar command that will internally start and commit transactions. In such a case, the original caller values of CurrentMemoryContext and CurrentResourceOwner will point to objects that will be destroyed by the internal commit. We must restore these pointers to point to the newly-manufactured transaction context and resource owner, rather than possibly pointing to deleted memory. Also tweak xact.c so that AbortTransaction and AbortSubTransaction forcibly restore a sane value for CurrentResourceOwner, much as they have always done for CurrentMemoryContext. I'm not certain this is necessary but I'm feeling paranoid today. Responds to Sean Chittenden's bug report of 4-Oct.
2004-10-04 23:52:15 +02:00
static void AtSubAbort_ResourceOwner(void);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void AtSubCommit_Memory(void);
static void AtSubStart_Memory(void);
static void AtSubStart_ResourceOwner(void);
static void ShowTransactionState(const char *str);
Harmonize parameter names in storage and AM code. Make sure that function declarations use names that exactly match the corresponding names from function definitions in storage, catalog, access method, executor, and logical replication code, as well as in miscellaneous utility/library code. Like other recent commits that cleaned up function parameter names, this commit was written with help from clang-tidy. Later commits will do the same for other parts of the codebase. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: David Rowley <dgrowleyml@gmail.com> Discussion: https://postgr.es/m/CAH2-WznJt9CMM9KJTMjJh_zbL5hD9oX44qdJ4aqZtjFi-zA3Tg@mail.gmail.com
2022-09-20 04:18:36 +02:00
static void ShowTransactionStateRec(const char *str, TransactionState s);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static const char *BlockStateAsString(TBlockState blockState);
static const char *TransStateAsString(TransState state);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* transaction state accessors
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
* ----------------------------------------------------------------
*/
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* IsTransactionState
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
Redefine IsTransactionState() to only return true for TRANS_INPROGRESS state, which is the only state in which it's safe to initiate database queries. It turns out that all but two of the callers thought that's what it meant; and the other two were using it as a proxy for "will GetTopTransactionId() return a nonzero XID"? Since it was in fact an unreliable guide to that, make those two just invoke GetTopTransactionId() always, then deal with a zero result if they get one.
2007-06-07 23:45:59 +02:00
* This returns true if we are inside a valid transaction; that is,
* it is safe to initiate database access, take heavyweight locks, etc.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
bool
All external function definitions now have prototypes that are checked.
1996-11-10 04:06:38 +01:00
IsTransactionState(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Redefine IsTransactionState() to only return true for TRANS_INPROGRESS state, which is the only state in which it's safe to initiate database queries. It turns out that all but two of the callers thought that's what it meant; and the other two were using it as a proxy for "will GetTopTransactionId() return a nonzero XID"? Since it was in fact an unreliable guide to that, make those two just invoke GetTopTransactionId() always, then deal with a zero result if they get one.
2007-06-07 23:45:59 +02:00
* TRANS_DEFAULT and TRANS_ABORT are obviously unsafe states. However, we
* also reject the startup/shutdown states TRANS_START, TRANS_COMMIT,
* TRANS_PREPARE since it might be too soon or too late within those
* transition states to do anything interesting. Hence, the only "valid"
* state is TRANS_INPROGRESS.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Redefine IsTransactionState() to only return true for TRANS_INPROGRESS state, which is the only state in which it's safe to initiate database queries. It turns out that all but two of the callers thought that's what it meant; and the other two were using it as a proxy for "will GetTopTransactionId() return a nonzero XID"? Since it was in fact an unreliable guide to that, make those two just invoke GetTopTransactionId() always, then deal with a zero result if they get one.
2007-06-07 23:45:59 +02:00
return (s->state == TRANS_INPROGRESS);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* IsAbortedTransactionBlockState
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
Teach standby conflict resolution to use SIGUSR1 Conflict reason is passed through directly to the backend, so we can take decisions about the effect of the conflict based upon the local state. No specific changes, as yet, though this prepares for later work. CancelVirtualTransaction() sends signals while holding ProcArrayLock. Introduce errdetail_abort() to give message detail explaining that the abort was caused by conflict processing. Remove CONFLICT_MODE states in favour of using PROCSIG_RECOVERY_CONFLICT states directly, for clarity.
2010-01-16 11:05:59 +01:00
* This returns true if we are within an aborted transaction block.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
bool
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
IsAbortedTransactionBlockState(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
if (s->blockState == TBLOCK_ABORT ||
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
s->blockState == TBLOCK_SUBABORT)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
return true;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
return false;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* GetTopTransactionId
*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* This will return the XID of the main transaction, assigning one if
* it's not yet set. Be careful to call this only inside a valid xact.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
TransactionId
GetTopTransactionId(void)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (!FullTransactionIdIsValid(XactTopFullTransactionId))
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
AssignTransactionId(&TopTransactionStateData);
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
return XidFromFullTransactionId(XactTopFullTransactionId);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* GetTopTransactionIdIfAny
*
* This will return the XID of the main transaction, if one is assigned.
* It will return InvalidTransactionId if we are not currently inside a
* transaction, or inside a transaction that hasn't yet been assigned an XID.
*/
TransactionId
GetTopTransactionIdIfAny(void)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
return XidFromFullTransactionId(XactTopFullTransactionId);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* GetCurrentTransactionId
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* This will return the XID of the current transaction (main or sub
* transaction), assigning one if it's not yet set. Be careful to call this
* only inside a valid xact.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
TransactionId
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
GetCurrentTransactionId(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (!FullTransactionIdIsValid(s->fullTransactionId))
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
AssignTransactionId(s);
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
return XidFromFullTransactionId(s->fullTransactionId);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
}
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* GetCurrentTransactionIdIfAny
*
* This will return the XID of the current sub xact, if one is assigned.
* It will return InvalidTransactionId if we are not currently inside a
* transaction, or inside a transaction that hasn't been assigned an XID yet.
*/
TransactionId
GetCurrentTransactionIdIfAny(void)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
return XidFromFullTransactionId(CurrentTransactionState->fullTransactionId);
}
/*
* GetTopFullTransactionId
*
* This will return the FullTransactionId of the main transaction, assigning
* one if it's not yet set. Be careful to call this only inside a valid xact.
*/
FullTransactionId
GetTopFullTransactionId(void)
{
if (!FullTransactionIdIsValid(XactTopFullTransactionId))
AssignTransactionId(&TopTransactionStateData);
return XactTopFullTransactionId;
}
/*
* GetTopFullTransactionIdIfAny
*
* This will return the FullTransactionId of the main transaction, if one is
* assigned. It will return InvalidFullTransactionId if we are not currently
* inside a transaction, or inside a transaction that hasn't yet been assigned
* one.
*/
FullTransactionId
GetTopFullTransactionIdIfAny(void)
{
return XactTopFullTransactionId;
}
/*
* GetCurrentFullTransactionId
*
* This will return the FullTransactionId of the current transaction (main or
* sub transaction), assigning one if it's not yet set. Be careful to call
* this only inside a valid xact.
*/
FullTransactionId
GetCurrentFullTransactionId(void)
{
TransactionState s = CurrentTransactionState;
if (!FullTransactionIdIsValid(s->fullTransactionId))
AssignTransactionId(s);
return s->fullTransactionId;
}
/*
* GetCurrentFullTransactionIdIfAny
*
* This will return the FullTransactionId of the current sub xact, if one is
* assigned. It will return InvalidFullTransactionId if we are not currently
* inside a transaction, or inside a transaction that hasn't been assigned one
* yet.
*/
FullTransactionId
GetCurrentFullTransactionIdIfAny(void)
{
return CurrentTransactionState->fullTransactionId;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
/*
* MarkCurrentTransactionIdLoggedIfAny
*
* Remember that the current xid - if it is assigned - now has been wal logged.
*/
void
MarkCurrentTransactionIdLoggedIfAny(void)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
CurrentTransactionState->didLogXid = true;
}
Replace XLOG_INCLUDE_XID flag with a more localized flag. Commit 0bead9af484c introduced XLOG_INCLUDE_XID flag to indicate that the WAL record contains subXID-to-topXID association. It uses that flag later to mark in CurrentTransactionState that top-xid is logged so that we should not try to log it again with the next WAL record in the current subtransaction. However, we can use a localized variable to pass that information. In passing, change the related function and variable names to make them consistent with what the code is actually doing. Author: Dilip Kumar Reviewed-by: Alvaro Herrera, Amit Kapila Discussion: https://postgr.es/m/E1mSoYz-0007Fh-D9@gemulon.postgresql.org
2021-11-02 04:05:29 +01:00
/*
* IsSubxactTopXidLogPending
*
* This is used to decide whether we need to WAL log the top-level XID for
* operation in a subtransaction. We require that for logical decoding, see
* LogicalDecodingProcessRecord.
*
* This returns true if wal_level >= logical and we are inside a valid
* subtransaction, for which the assignment was not yet written to any WAL
* record.
*/
bool
IsSubxactTopXidLogPending(void)
{
/* check whether it is already logged */
if (CurrentTransactionState->topXidLogged)
return false;
/* wal_level has to be logical */
if (!XLogLogicalInfoActive())
return false;
/* we need to be in a transaction state */
if (!IsTransactionState())
return false;
/* it has to be a subtransaction */
if (!IsSubTransaction())
return false;
/* the subtransaction has to have a XID assigned */
if (!TransactionIdIsValid(GetCurrentTransactionIdIfAny()))
return false;
return true;
}
/*
* MarkSubxactTopXidLogged
*
* Remember that the top transaction id for the current subtransaction is WAL
* logged now.
*/
void
MarkSubxactTopXidLogged(void)
{
Assert(IsSubxactTopXidLogPending());
CurrentTransactionState->topXidLogged = true;
}
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
/*
Ensure backwards compatibility for GetStableLatestTransactionId()
2012-05-12 14:26:10 +02:00
* GetStableLatestTransactionId
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
*
Improve comment for GetStableLatestTransactionId().
2012-05-31 17:20:02 +02:00
* Get the transaction's XID if it has one, else read the next-to-be-assigned
* XID. Once we have a value, return that same value for the remainder of the
* current transaction. This is meant to provide the reference point for the
* age(xid) function, but might be useful for other maintenance tasks as well.
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
*/
TransactionId
GetStableLatestTransactionId(void)
{
static LocalTransactionId lxid = InvalidLocalTransactionId;
static TransactionId stablexid = InvalidTransactionId;
Ensure backwards compatibility for GetStableLatestTransactionId()
2012-05-12 14:26:10 +02:00
if (lxid != MyProc->lxid)
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
{
lxid = MyProc->lxid;
Ensure backwards compatibility for GetStableLatestTransactionId()
2012-05-12 14:26:10 +02:00
stablexid = GetTopTransactionIdIfAny();
if (!TransactionIdIsValid(stablexid))
ReadNewTransactionId() -> ReadNextTransactionId(). The new name conveys the effect better, is more consistent with similar functions ReadNextMultiXactId(), ReadNextFullTransactionId(), and matches the name of the variable that it reads. Reported-by: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/CAH2-WzmVR4SakBXQUdhhPpMf1aYvZCnna5%3DHKa7DAgEmBAg%2B8g%40mail.gmail.com
2021-02-15 01:03:10 +01:00
stablexid = ReadNextTransactionId();
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
}
Ensure backwards compatibility for GetStableLatestTransactionId()
2012-05-12 14:26:10 +02:00
Assert(TransactionIdIsValid(stablexid));
Ensure age() returns a stable value rather than the latest value
2012-05-11 15:36:24 +02:00
return stablexid;
}
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* AssignTransactionId
*
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
* Assigns a new permanent FullTransactionId to the given TransactionState.
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* We do not assign XIDs to transactions until/unless this is called.
* Also, any parent TransactionStates that don't yet have XIDs are assigned
* one; this maintains the invariant that a child transaction has an XID
* following its parent's.
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
static void
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
AssignTransactionId(TransactionState s)
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
{
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
bool isSubXact = (s->parent != NULL);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
ResourceOwner currentOwner;
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
bool log_unknown_top = false;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Assert that caller didn't screw up */
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
Assert(!FullTransactionIdIsValid(s->fullTransactionId));
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Assert(s->state == TRANS_INPROGRESS);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for new XIDs at this point.
*/
Tighten up application of parallel mode checks. Commit 924bcf4f16d54c55310b28f77686608684734f42 failed to enforce parallel mode checks during the commit of a parallel worker, because we exited parallel mode prior to ending the transaction so that we could pop the active snapshot. Re-establish parallel mode during parallel worker commit. Without this, it's far too easy for unsafe actions during the pre-commit sequence to crash the server instead of hitting the error checks as intended. Just to be extra paranoid, adjust a couple of the sanity checks in xact.c to check not only IsInParallelMode() but also IsParallelWorker().
2015-10-16 15:59:57 +02:00
if (IsInParallelMode() || IsParallelWorker())
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
elog(ERROR, "cannot assign XIDs during a parallel operation");
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* Ensure parent(s) have XIDs, so that a child always has an XID later
Fix spelling errors and typos in comments Author: Daniel Gustafsson <daniel@yesql.se>
2018-11-02 13:56:16 +01:00
* than its parent. Mustn't recurse here, or we might get a stack
Avoid deep recursion when assigning XIDs to multiple levels of subxacts. Backpatch to 8.0. Andres Freund, with cleanup and adjustment for older branches by me.
2010-07-23 02:43:00 +02:00
* overflow if we're at the bottom of a huge stack of subtransactions none
* of which have XIDs yet.
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
*/
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (isSubXact && !FullTransactionIdIsValid(s->parent->fullTransactionId))
Avoid deep recursion when assigning XIDs to multiple levels of subxacts. Backpatch to 8.0. Andres Freund, with cleanup and adjustment for older branches by me.
2010-07-23 02:43:00 +02:00
{
TransactionState p = s->parent;
TransactionState *parents;
size_t parentOffset = 0;
parents = palloc(sizeof(TransactionState) * s->nestingLevel);
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
while (p != NULL && !FullTransactionIdIsValid(p->fullTransactionId))
Avoid deep recursion when assigning XIDs to multiple levels of subxacts. Backpatch to 8.0. Andres Freund, with cleanup and adjustment for older branches by me.
2010-07-23 02:43:00 +02:00
{
parents[parentOffset++] = p;
p = p->parent;
}
/*
* This is technically a recursive call, but the recursion will never
* be more than one layer deep.
*/
while (parentOffset != 0)
AssignTransactionId(parents[--parentOffset]);
pfree(parents);
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
/*
* When wal_level=logical, guarantee that a subtransaction's xid can only
* be seen in the WAL stream if its toplevel xid has been logged before.
Collection of typo fixes. Use "a" and "an" correctly, mostly in comments. Two error messages were also fixed (they were just elogs, so no translation work required). Two function comments in pg_proc.h were also fixed. Etsuro Fujita reported one of these, but I found a lot more with grep. Also fix a few other typos spotted while grepping for the a/an typos. For example, "consists out of ..." -> "consists of ...". Plus a "though"/ "through" mixup reported by Euler Taveira. Many of these typos were in old code, which would be nice to backpatch to make future backpatching easier. But much of the code was new, and I didn't feel like crafting separate patches for each branch. So no backpatching.
2015-05-20 15:18:11 +02:00
* If necessary we log an xact_assignment record with fewer than
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
* PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
* for a transaction even though it appears in a WAL record, we just might
* superfluously log something. That can happen when an xid is included
* somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
* xl_standby_locks.
*/
if (isSubXact && XLogLogicalInfoActive() &&
!TopTransactionStateData.didLogXid)
log_unknown_top = true;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
Fix typos referring to PGPROC Japin Li Reviewed by Kyotaro Horiguchi Discussion: https://www.postgresql.org/message-id/MEYP282MB1669459813B36FB5EAA38434B6499@MEYP282MB1669.AUSP282.PROD.OUTLOOK.COM
2022-09-19 06:36:51 +02:00
* Generate a new FullTransactionId and record its xid in PGPROC and
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
* pg_subtrans.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* NB: we must make the subtrans entry BEFORE the Xid appears anywhere in
Fix typos referring to PGPROC Japin Li Reviewed by Kyotaro Horiguchi Discussion: https://www.postgresql.org/message-id/MEYP282MB1669459813B36FB5EAA38434B6499@MEYP282MB1669.AUSP282.PROD.OUTLOOK.COM
2022-09-19 06:36:51 +02:00
* shared storage other than PGPROC; because if there's no room for it in
* PGPROC, the subtrans entry is needed to ensure that other backends see
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* the Xid as "running". See GetNewTransactionId.
*/
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = GetNewTransactionId(isSubXact);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
if (!isSubXact)
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
XactTopFullTransactionId = s->fullTransactionId;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
if (isSubXact)
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
SubTransSetParent(XidFromFullTransactionId(s->fullTransactionId),
XidFromFullTransactionId(s->parent->fullTransactionId));
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Move RegisterPredicateLockingXid() call to a safer place. The SSI patch inserted a call of RegisterPredicateLockingXid into GetNewTransactionId, which was a bad idea on a couple of grounds. First, it's not necessary to hold XidGenLock while manipulating that shared memory, and doing so is bad because XidGenLock is a high-contention lock that should be held for as short a time as possible. (Not to mention that it adds an entirely unnecessary deadlock hazard, since we must take SerializableXactHashLock as well.) Second, the specific place where it was put was between extending CLOG and advancing nextXid, which could result in unpleasant behavior in case of a failure there. Pull the call out to AssignTransactionId, which is much safer and arguably better from a modularity standpoint too. There is more work to do to clean up the failure-before-advancing-nextXid issue, but that is a separate change that will need to be back-patched. So for the moment I just want to make GetNewTransactionId look the same as it did in prior versions.
2011-05-06 18:57:28 +02:00
/*
* If it's a top-level transaction, the predicate locking system needs to
* be told about it too.
*/
if (!isSubXact)
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
RegisterPredicateLockingXid(XidFromFullTransactionId(s->fullTransactionId));
Move RegisterPredicateLockingXid() call to a safer place. The SSI patch inserted a call of RegisterPredicateLockingXid into GetNewTransactionId, which was a bad idea on a couple of grounds. First, it's not necessary to hold XidGenLock while manipulating that shared memory, and doing so is bad because XidGenLock is a high-contention lock that should be held for as short a time as possible. (Not to mention that it adds an entirely unnecessary deadlock hazard, since we must take SerializableXactHashLock as well.) Second, the specific place where it was put was between extending CLOG and advancing nextXid, which could result in unpleasant behavior in case of a failure there. Pull the call out to AssignTransactionId, which is much safer and arguably better from a modularity standpoint too. There is more work to do to clean up the failure-before-advancing-nextXid issue, but that is a separate change that will need to be back-patched. So for the moment I just want to make GetNewTransactionId look the same as it did in prior versions.
2011-05-06 18:57:28 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* Acquire lock on the transaction XID. (We assume this cannot block.) We
* have to ensure that the lock is assigned to the transaction's own
* ResourceOwner.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*/
currentOwner = CurrentResourceOwner;
Remove unnecessary PG_TRY overhead for CurrentResourceOwner changes. resowner/README contained advice to use a PG_TRY block to restore the old CurrentResourceOwner value anywhere that that variable is transiently changed. That advice was only inconsistently followed, however, and on reflection it seems like unnecessary overhead. We don't bother with such a convention for transient CurrentMemoryContext changes, on the grounds that any (sub)transaction abort will start out by resetting CurrentMemoryContext to what it wants. But the same is true of CurrentResourceOwner, so there seems no need to treat it differently. Hence, remove PG_TRY blocks that exist only to restore CurrentResourceOwner before re-throwing the error. There are a couple of places that restore it along with some other actions, and I left those alone; the restore is probably unnecessary but no noticeable gain will result from removing it. Discussion: https://postgr.es/m/5236.1507583529@sss.pgh.pa.us
2017-10-11 23:43:50 +02:00
CurrentResourceOwner = s->curTransactionOwner;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
XactLockTableInsert(XidFromFullTransactionId(s->fullTransactionId));
Remove unnecessary PG_TRY overhead for CurrentResourceOwner changes. resowner/README contained advice to use a PG_TRY block to restore the old CurrentResourceOwner value anywhere that that variable is transiently changed. That advice was only inconsistently followed, however, and on reflection it seems like unnecessary overhead. We don't bother with such a convention for transient CurrentMemoryContext changes, on the grounds that any (sub)transaction abort will start out by resetting CurrentMemoryContext to what it wants. But the same is true of CurrentResourceOwner, so there seems no need to treat it differently. Hence, remove PG_TRY blocks that exist only to restore CurrentResourceOwner before re-throwing the error. There are a couple of places that restore it along with some other actions, and I left those alone; the restore is probably unnecessary but no noticeable gain will result from removing it. Discussion: https://postgr.es/m/5236.1507583529@sss.pgh.pa.us
2017-10-11 23:43:50 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CurrentResourceOwner = currentOwner;
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* Every PGPROC_MAX_CACHED_SUBXIDS assigned transaction ids within each
* top-level transaction we issue a WAL record for the assignment. We
* include the top-level xid and all the subxids that have not yet been
* reported using XLOG_XACT_ASSIGNMENT records.
*
* This is required to limit the amount of shared memory required in a hot
* standby server to keep track of in-progress XIDs. See notes for
* RecordKnownAssignedTransactionIds().
*
* We don't keep track of the immediate parent of each subxid, only the
* top-level transaction that each subxact belongs to. This is correct in
* recovery only because aborted subtransactions are separately WAL
* logged.
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
*
* This is correct even for the case where several levels above us didn't
* have an xid assigned as we recursed up to them beforehand.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
if (isSubXact && XLogStandbyInfoActive())
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
unreportedXids[nUnreportedXids] = XidFromFullTransactionId(s->fullTransactionId);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
nUnreportedXids++;
/*
* ensure this test matches similar one in
* RecoverPreparedTransactions()
*/
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
if (nUnreportedXids >= PGPROC_MAX_CACHED_SUBXIDS ||
log_unknown_top)
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
{
xl_xact_assignment xlrec;
/*
* xtop is always set by now because we recurse up transaction
* stack to the highest unassigned xid and then come back down
*/
xlrec.xtop = GetTopTransactionId();
Assert(TransactionIdIsValid(xlrec.xtop));
xlrec.nsubxacts = nUnreportedXids;
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
XLogRegisterData((char *) unreportedXids,
nUnreportedXids * sizeof(TransactionId));
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
nUnreportedXids = 0;
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
/* mark top, not current xact as having been logged */
TopTransactionStateData.didLogXid = true;
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
}
}
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* GetCurrentSubTransactionId
*/
SubTransactionId
GetCurrentSubTransactionId(void)
{
TransactionState s = CurrentTransactionState;
return s->subTransactionId;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Fix SQL function execution to be safe with long-lived FmgrInfos. fmgr_sql had been designed on the assumption that the FmgrInfo it's called with has only query lifespan. This is demonstrably unsafe in connection with range types, as shown in bug #7881 from Andrew Gierth. Fix things so that we re-generate the function's cache data if the (sub)transaction it was made in is no longer active. Back-patch to 9.2. This might be needed further back, but it's not clear whether the case can realistically arise without range types, so for now I'll desist from back-patching further.
2013-03-03 23:39:58 +01:00
/*
* SubTransactionIsActive
*
* Test if the specified subxact ID is still active. Note caller is
* responsible for checking whether this ID is relevant to the current xact.
*/
bool
SubTransactionIsActive(SubTransactionId subxid)
{
TransactionState s;
for (s = CurrentTransactionState; s != NULL; s = s->parent)
{
if (s->state == TRANS_ABORT)
continue;
if (s->subTransactionId == subxid)
return true;
}
return false;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* GetCurrentCommandId
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
*
Change TRUE/FALSE to true/false The lower case spellings are C and C++ standard and are used in most parts of the PostgreSQL sources. The upper case spellings are only used in some files/modules. So standardize on the standard spellings. The APIs for ICU, Perl, and Windows define their own TRUE and FALSE, so those are left as is when using those APIs. In code comments, we use the lower-case spelling for the C concepts and keep the upper-case spelling for the SQL concepts. Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
2017-08-16 06:22:32 +02:00
* "used" must be true if the caller intends to use the command ID to mark
* inserted/updated/deleted tuples. false means the ID is being fetched
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
* for read-only purposes (ie, as a snapshot validity cutoff). See
* CommandCounterIncrement() for discussion.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
CommandId
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
GetCurrentCommandId(bool used)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* this is global to a transaction, not subtransaction-local */
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
if (used)
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
{
/*
Adjust assertion in GetCurrentCommandId. currentCommandIdUsed is only used to skip redundant increments of the command counter, and CommandCounterIncrement() is categorically denied under parallelism anyway. Therefore, it's OK for GetCurrentCommandId() to mark the counter value used, as long as it happens in the leader, not a worker. Prior to commit e9baa5e9fa147e00a2466ab2c40eb99c8a700824, the slightly incorrect check didn't matter, but now it does. A test case added by commit 1804284042e659e7d16904e7bbb0ad546394b6a3 uncovered the problem by accident; it caused failures with force_parallel_mode=on/regress. Report and review by Andres Freund. Patch by me. Discussion: http://postgr.es/m/20171221143106.5lhtygohvmazli3x@alap3.anarazel.de
2017-12-21 19:10:51 +01:00
* Forbid setting currentCommandIdUsed in a parallel worker, because
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* we have no provision for communicating this back to the leader. We
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* could relax this restriction when currentCommandIdUsed was already
* true at the start of the parallel operation.
*/
Adjust assertion in GetCurrentCommandId. currentCommandIdUsed is only used to skip redundant increments of the command counter, and CommandCounterIncrement() is categorically denied under parallelism anyway. Therefore, it's OK for GetCurrentCommandId() to mark the counter value used, as long as it happens in the leader, not a worker. Prior to commit e9baa5e9fa147e00a2466ab2c40eb99c8a700824, the slightly incorrect check didn't matter, but now it does. A test case added by commit 1804284042e659e7d16904e7bbb0ad546394b6a3 uncovered the problem by accident; it caused failures with force_parallel_mode=on/regress. Report and review by Andres Freund. Patch by me. Discussion: http://postgr.es/m/20171221143106.5lhtygohvmazli3x@alap3.anarazel.de
2017-12-21 19:10:51 +01:00
Assert(!IsParallelWorker());
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
currentCommandIdUsed = true;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
return currentCommandId;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
/*
* SetParallelStartTimestamps
*
* In a parallel worker, we should inherit the parent transaction's
* timestamps rather than setting our own. The parallel worker
* infrastructure must call this to provide those values before
* calling StartTransaction() or SetCurrentStatementStartTimestamp().
*/
void
SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts)
{
Assert(IsParallelWorker());
xactStartTimestamp = xact_ts;
stmtStartTimestamp = stmt_ts;
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Clean up the rather historically encumbered interface to now() and current time: provide a GetCurrentTimestamp() function that returns current time in the form of a TimestampTz, instead of separate time_t and microseconds fields. This is what all the callers really want anyway, and it eliminates low-level dependencies on AbsoluteTime, which is a deprecated datatype that will have to disappear eventually.
2005-06-30 00:51:57 +02:00
* GetCurrentTransactionStartTimestamp
Measure the current transaction time to milliseconds. Define a new function, GetCurrentTransactionStartTimeUsec() to get the time to this precision. Allow now() and timestamp 'now' to use this higher precision result so we now have fractional seconds in this "constant". Add timestamp without time zone type. Move previous timestamp type to timestamp with time zone. Accept another ISO variant for date/time values: yyyy-mm-ddThh:mm:ss (note the "T" separating the day from hours information). Remove 'current' from date/time types; convert to 'now' in input. Separate time and timetz regression tests. Separate timestamp and timestamptz regression test.
2001-09-28 10:09:14 +02:00
*/
Clean up the rather historically encumbered interface to now() and current time: provide a GetCurrentTimestamp() function that returns current time in the form of a TimestampTz, instead of separate time_t and microseconds fields. This is what all the callers really want anyway, and it eliminates low-level dependencies on AbsoluteTime, which is a deprecated datatype that will have to disappear eventually.
2005-06-30 00:51:57 +02:00
TimestampTz
GetCurrentTransactionStartTimestamp(void)
Measure the current transaction time to milliseconds. Define a new function, GetCurrentTransactionStartTimeUsec() to get the time to this precision. Allow now() and timestamp 'now' to use this higher precision result so we now have fractional seconds in this "constant". Add timestamp without time zone type. Move previous timestamp type to timestamp with time zone. Accept another ISO variant for date/time values: yyyy-mm-ddThh:mm:ss (note the "T" separating the day from hours information). Remove 'current' from date/time types; convert to 'now' in input. Separate time and timetz regression tests. Separate timestamp and timestamptz regression test.
2001-09-28 10:09:14 +02:00
{
Clean up the rather historically encumbered interface to now() and current time: provide a GetCurrentTimestamp() function that returns current time in the form of a TimestampTz, instead of separate time_t and microseconds fields. This is what all the callers really want anyway, and it eliminates low-level dependencies on AbsoluteTime, which is a deprecated datatype that will have to disappear eventually.
2005-06-30 00:51:57 +02:00
return xactStartTimestamp;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
Add statement_timestamp(), clock_timestamp(), and transaction_timestamp() (just like now()). Also update statement_timeout() to mention it is statement arrival time that is measured. Catalog version updated.
2006-04-25 02:25:22 +02:00
/*
* GetCurrentStatementStartTimestamp
*/
TimestampTz
GetCurrentStatementStartTimestamp(void)
{
return stmtStartTimestamp;
}
Implement rate-limiting logic on how often backends will attempt to send messages to the stats collector. This avoids the problem that enabling stats_row_level for autovacuum has a significant overhead for short read-only transactions, as noted by Arjen van der Meijden. We can avoid an extra gettimeofday call by piggybacking on the one done for WAL-logging xact commit or abort (although that doesn't help read-only transactions, since they don't WAL-log anything). In my proposal for this, I noted that we could change the WAL log entries for commit/abort to record full TimestampTz precision, instead of only time_t as at present. That's not done in this patch, but will be committed separately.
2007-04-30 05:23:49 +02:00
/*
* GetCurrentTransactionStopTimestamp
*
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
* If the transaction stop time hasn't already been set, which can happen if
* we decided we don't need to log an XLOG record, set xactStopTimestamp.
Implement rate-limiting logic on how often backends will attempt to send messages to the stats collector. This avoids the problem that enabling stats_row_level for autovacuum has a significant overhead for short read-only transactions, as noted by Arjen van der Meijden. We can avoid an extra gettimeofday call by piggybacking on the one done for WAL-logging xact commit or abort (although that doesn't help read-only transactions, since they don't WAL-log anything). In my proposal for this, I noted that we could change the WAL log entries for commit/abort to record full TimestampTz precision, instead of only time_t as at present. That's not done in this patch, but will be committed separately.
2007-04-30 05:23:49 +02:00
*/
TimestampTz
GetCurrentTransactionStopTimestamp(void)
{
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
TransactionState s PG_USED_FOR_ASSERTS_ONLY = CurrentTransactionState;
/* should only be called after commit / abort processing */
Assert(s->state == TRANS_DEFAULT ||
s->state == TRANS_COMMIT ||
s->state == TRANS_ABORT ||
s->state == TRANS_PREPARE);
if (xactStopTimestamp == 0)
xactStopTimestamp = GetCurrentTimestamp();
return xactStopTimestamp;
Implement rate-limiting logic on how often backends will attempt to send messages to the stats collector. This avoids the problem that enabling stats_row_level for autovacuum has a significant overhead for short read-only transactions, as noted by Arjen van der Meijden. We can avoid an extra gettimeofday call by piggybacking on the one done for WAL-logging xact commit or abort (although that doesn't help read-only transactions, since they don't WAL-log anything). In my proposal for this, I noted that we could change the WAL log entries for commit/abort to record full TimestampTz precision, instead of only time_t as at present. That's not done in this patch, but will be committed separately.
2007-04-30 05:23:49 +02:00
}
Add statement_timestamp(), clock_timestamp(), and transaction_timestamp() (just like now()). Also update statement_timeout() to mention it is statement arrival time that is measured. Catalog version updated.
2006-04-25 02:25:22 +02:00
/*
* SetCurrentStatementStartTimestamp
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
*
* In a parallel worker, this should already have been provided by a call
* to SetParallelStartTimestamps().
Add statement_timestamp(), clock_timestamp(), and transaction_timestamp() (just like now()). Also update statement_timeout() to mention it is statement arrival time that is measured. Catalog version updated.
2006-04-25 02:25:22 +02:00
*/
void
SetCurrentStatementStartTimestamp(void)
{
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
if (!IsParallelWorker())
stmtStartTimestamp = GetCurrentTimestamp();
else
Assert(stmtStartTimestamp != 0);
Add statement_timestamp(), clock_timestamp(), and transaction_timestamp() (just like now()). Also update statement_timeout() to mention it is statement arrival time that is measured. Catalog version updated.
2006-04-25 02:25:22 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* GetCurrentTransactionNestLevel
*
* Note: this will return zero when not inside any transaction, one when
* inside a top-level transaction, etc.
*/
int
GetCurrentTransactionNestLevel(void)
{
TransactionState s = CurrentTransactionState;
Measure the current transaction time to milliseconds. Define a new function, GetCurrentTransactionStartTimeUsec() to get the time to this precision. Allow now() and timestamp 'now' to use this higher precision result so we now have fractional seconds in this "constant". Add timestamp without time zone type. Move previous timestamp type to timestamp with time zone. Accept another ISO variant for date/time values: yyyy-mm-ddThh:mm:ss (note the "T" separating the day from hours information). Remove 'current' from date/time types; convert to 'now' in input. Separate time and timetz regression tests. Separate timestamp and timestamptz regression test.
2001-09-28 10:09:14 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
return s->nestingLevel;
Measure the current transaction time to milliseconds. Define a new function, GetCurrentTransactionStartTimeUsec() to get the time to this precision. Allow now() and timestamp 'now' to use this higher precision result so we now have fractional seconds in this "constant". Add timestamp without time zone type. Move previous timestamp type to timestamp with time zone. Accept another ISO variant for date/time values: yyyy-mm-ddThh:mm:ss (note the "T" separating the day from hours information). Remove 'current' from date/time types; convert to 'now' in input. Separate time and timetz regression tests. Separate timestamp and timestamptz regression test.
2001-09-28 10:09:14 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* TransactionIdIsCurrentTransactionId
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
bool
TransactionIdIsCurrentTransactionId(TransactionId xid)
{
Fix relcache to account properly for subtransaction status of 'new' relcache entries. Also, change TransactionIdIsCurrentTransactionId() so that if consulted during transaction abort, it will not say that the aborted xact is still current. (It would be better to ensure that it's never called at all during abort, but I'm not sure we can easily guarantee that.) In combination, these fix a crash we have seen occasionally during parallel regression tests of 8.0.
2004-08-28 22:31:44 +02:00
TransactionState s;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Remove some no-longer-needed kluges for bootstrapping, in particular the AMI_OVERRIDE flag. The fact that TransactionLogFetch treats BootstrapTransactionId as always committed is sufficient to make bootstrap work, and getting rid of extra tests in heavily used code paths seems like a win. The files produced by initdb are demonstrably the same after this change.
2005-02-20 22:46:50 +01:00
/*
* We always say that BootstrapTransactionId is "not my transaction ID"
* even when it is (ie, during bootstrap). Along with the fact that
* transam.c always treats BootstrapTransactionId as already committed,
Move remaining code from tqual.[ch] to heapam.h / heapam_visibility.c. Given these routines are heap specific, and that there will be more generic visibility support in via table AM, it makes sense to move the prototypes to heapam.h (routines like HeapTupleSatisfiesVacuum will not be exposed in a generic fashion, because they are too storage specific). Similarly, the code in tqual.c is specific to heap, so moving it into access/heap/ makes sense. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-01-22 02:03:15 +01:00
* this causes the heapam_visibility.c routines to see all tuples as
* committed, which is what we need during bootstrap. (Bootstrap mode
* only inserts tuples, it never updates or deletes them, so all tuples
* can be presumed good immediately.)
Fix recently-understood problems with handling of XID freezing, particularly in PITR scenarios. We now WAL-log the replacement of old XIDs with FrozenTransactionId, so that such replacement is guaranteed to propagate to PITR slave databases. Also, rather than relying on hint-bit updates to be preserved, pg_clog is not truncated until all instances of an XID are known to have been replaced by FrozenTransactionId. Add new GUC variables and pg_autovacuum columns to allow management of the freezing policy, so that users can trade off the size of pg_clog against the amount of freezing work done. Revise the already-existing code that forces autovacuum of tables approaching the wraparound point to make it more bulletproof; also, revise the autovacuum logic so that anti-wraparound vacuuming is done per-table rather than per-database. initdb forced because of changes in pg_class, pg_database, and pg_autovacuum catalogs. Heikki Linnakangas, Simon Riggs, and Tom Lane.
2006-11-05 23:42:10 +01:00
*
* Likewise, InvalidTransactionId and FrozenTransactionId are certainly
* not my transaction ID, so we can just return "false" immediately for
* any non-normal XID.
Remove some no-longer-needed kluges for bootstrapping, in particular the AMI_OVERRIDE flag. The fact that TransactionLogFetch treats BootstrapTransactionId as always committed is sufficient to make bootstrap work, and getting rid of extra tests in heavily used code paths seems like a win. The files produced by initdb are demonstrably the same after this change.
2005-02-20 22:46:50 +01:00
*/
Fix recently-understood problems with handling of XID freezing, particularly in PITR scenarios. We now WAL-log the replacement of old XIDs with FrozenTransactionId, so that such replacement is guaranteed to propagate to PITR slave databases. Also, rather than relying on hint-bit updates to be preserved, pg_clog is not truncated until all instances of an XID are known to have been replaced by FrozenTransactionId. Add new GUC variables and pg_autovacuum columns to allow management of the freezing policy, so that users can trade off the size of pg_clog against the amount of freezing work done. Revise the already-existing code that forces autovacuum of tables approaching the wraparound point to make it more bulletproof; also, revise the autovacuum logic so that anti-wraparound vacuuming is done per-table rather than per-database. initdb forced because of changes in pg_class, pg_database, and pg_autovacuum catalogs. Heikki Linnakangas, Simon Riggs, and Tom Lane.
2006-11-05 23:42:10 +01:00
if (!TransactionIdIsNormal(xid))
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
return false;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Optimize TransactionIdIsCurrentTransactionId(). If the passed in xid is the current top transaction, we can do a fast check and exit early. This should work well for the current heap but also works very well for proposed AMs that don't use a separate xid for subtransactions. Author: Ashwin Agrawal, based on a suggestion from Andres Freund Reviewed-by: Thomas Munro Discussion: https://postgr.es/m/CALfoeiv0k3hkEb3Oqk%3DziWqtyk2Jys1UOK5hwRBNeANT_yX%2Bng%40mail.gmail.com
2019-11-11 04:33:04 +01:00
if (TransactionIdEquals(xid, GetTopTransactionIdIfAny()))
return true;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* In parallel workers, the XIDs we must consider as current are stored in
* ParallelCurrentXids rather than the transaction-state stack. Note that
* the XIDs in this array are sorted numerically rather than according to
* transactionIdPrecedes order.
*/
if (nParallelCurrentXids > 0)
{
int low,
high;
low = 0;
high = nParallelCurrentXids - 1;
while (low <= high)
{
int middle;
TransactionId probe;
middle = low + (high - low) / 2;
probe = ParallelCurrentXids[middle];
if (probe == xid)
return true;
else if (probe < xid)
low = middle + 1;
else
high = middle - 1;
}
return false;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* We will return true for the Xid of the current subtransaction, any of
* its subcommitted children, any of its parents, or any of their
Fix relcache to account properly for subtransaction status of 'new' relcache entries. Also, change TransactionIdIsCurrentTransactionId() so that if consulted during transaction abort, it will not say that the aborted xact is still current. (It would be better to ensure that it's never called at all during abort, but I'm not sure we can easily guarantee that.) In combination, these fix a crash we have seen occasionally during parallel regression tests of 8.0.
2004-08-28 22:31:44 +02:00
* previously subcommitted children. However, a transaction being aborted
* is no longer "current", even though it may still have an entry on the
* state stack.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Fix relcache to account properly for subtransaction status of 'new' relcache entries. Also, change TransactionIdIsCurrentTransactionId() so that if consulted during transaction abort, it will not say that the aborted xact is still current. (It would be better to ensure that it's never called at all during abort, but I'm not sure we can easily guarantee that.) In combination, these fix a crash we have seen occasionally during parallel regression tests of 8.0.
2004-08-28 22:31:44 +02:00
for (s = CurrentTransactionState; s != NULL; s = s->parent)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
{
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
int low,
high;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Fix relcache to account properly for subtransaction status of 'new' relcache entries. Also, change TransactionIdIsCurrentTransactionId() so that if consulted during transaction abort, it will not say that the aborted xact is still current. (It would be better to ensure that it's never called at all during abort, but I'm not sure we can easily guarantee that.) In combination, these fix a crash we have seen occasionally during parallel regression tests of 8.0.
2004-08-28 22:31:44 +02:00
if (s->state == TRANS_ABORT)
continue;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (!FullTransactionIdIsValid(s->fullTransactionId))
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
continue; /* it can't have any child XIDs either */
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (TransactionIdEquals(xid, XidFromFullTransactionId(s->fullTransactionId)))
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
return true;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
/* As the childXids array is ordered, we can use binary search */
low = 0;
high = s->nChildXids - 1;
while (low <= high)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
{
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
int middle;
TransactionId probe;
middle = low + (high - low) / 2;
probe = s->childXids[middle];
if (TransactionIdEquals(probe, xid))
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
return true;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
else if (TransactionIdPrecedes(probe, xid))
low = middle + 1;
else
high = middle - 1;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
}
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
return false;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* TransactionStartedDuringRecovery
*
* Returns true if the current transaction started while recovery was still
* in progress. Recovery might have ended since so RecoveryInProgress() might
* return false already.
*/
bool
TransactionStartedDuringRecovery(void)
{
return CurrentTransactionState->startedInRecovery;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* EnterParallelMode
*/
void
EnterParallelMode(void)
{
TransactionState s = CurrentTransactionState;
Assert(s->parallelModeLevel >= 0);
++s->parallelModeLevel;
}
/*
* ExitParallelMode
*/
void
ExitParallelMode(void)
{
TransactionState s = CurrentTransactionState;
Assert(s->parallelModeLevel > 0);
Assert(s->parallelModeLevel > 1 || !ParallelContextActive());
--s->parallelModeLevel;
}
/*
* IsInParallelMode
*
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* Are we in a parallel operation, as either the leader or a worker? Check
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* this to prohibit operations that change backend-local state expected to
* match across all workers. Mere caches usually don't require such a
* restriction. State modified in a strict push/pop fashion, such as the
* active snapshot stack, is often fine.
*/
bool
IsInParallelMode(void)
{
return CurrentTransactionState->parallelModeLevel != 0;
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* CommandCounterIncrement
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
CommandCounterIncrement(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
/*
* If the current value of the command counter hasn't been "used" to mark
* tuples, we need not increment it, since there's no need to distinguish
* a read-only command from others. This helps postpone command counter
* overflow, and keeps no-op CommandCounterIncrement operations cheap.
*/
if (currentCommandIdUsed)
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
{
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each
* parallel operation, so we can't account for new commands after that
* point.
*/
Tighten up application of parallel mode checks. Commit 924bcf4f16d54c55310b28f77686608684734f42 failed to enforce parallel mode checks during the commit of a parallel worker, because we exited parallel mode prior to ending the transaction so that we could pop the active snapshot. Re-establish parallel mode during parallel worker commit. Without this, it's far too easy for unsafe actions during the pre-commit sequence to crash the server instead of hitting the error checks as intended. Just to be extra paranoid, adjust a couple of the sanity checks in xact.c to check not only IsInParallelMode() but also IsParallelWorker().
2015-10-16 15:59:57 +02:00
if (IsInParallelMode() || IsParallelWorker())
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
elog(ERROR, "cannot start commands during a parallel operation");
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
currentCommandId += 1;
Introduce InvalidCommandId. This allows a 32-bit field to represent an *optional* command ID without a separate flag bit. Andres Freund
2013-09-09 22:25:29 +02:00
if (currentCommandId == InvalidCommandId)
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
{
currentCommandId -= 1;
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
Introduce InvalidCommandId. This allows a 32-bit field to represent an *optional* command ID without a separate flag bit. Andres Freund
2013-09-09 22:25:29 +02:00
errmsg("cannot have more than 2^32-2 commands in a transaction")));
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
}
currentCommandIdUsed = false;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Improve snapshot manager by keeping explicit track of snapshots. There are two ways to track a snapshot: there's the "registered" list, which is used for arbitrary long-lived snapshots; and there's the "active stack", which is used for the snapshot that is considered "active" at any time. This also allows users of snapshots to stop worrying about snapshot memory allocation and freeing, and about using PG_TRY blocks around ActiveSnapshot assignment. This is all done automatically now. As a consequence, this allows us to reset MyProc->xmin when there are no more snapshots registered in the current backend, reducing the impact that long-running transactions have on VACUUM.
2008-05-12 22:02:02 +02:00
/* Propagate new command ID into static snapshots */
SnapshotSetCommandId(currentCommandId);
8.4 pgindent run, with new combined Linux/FreeBSD/MinGW typedef list provided by Andrew.
2009-06-11 16:49:15 +02:00
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
/*
* Make any catalog changes done by the just-completed command visible
* in the local syscache. We obviously don't need to do this after a
* read-only command. (But see hacks in inval.c to make real sure we
* don't think a command that queued inval messages was read-only.)
*/
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
AtCCI_LocalCache();
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
/*
* ForceSyncCommit
*
* Interface routine to allow commands to force a synchronous commit of the
Remove dead forceSync parameter of XactLogCommitRecord(). The function has been reading global variable forceSyncCommit, mirroring the intent of the caller that passed forceSync=forceSyncCommit. The other caller, RecordTransactionCommitPrepared(), passed false. Since COMMIT PREPARED can't share a transaction with any command, it certainly doesn't share a transaction with a command that sets forceSyncCommit. Reviewed by Michael Paquier. Discussion: https://postgr.es/m/20200617032615.GC2916904@rfd.leadboat.com
2020-06-20 10:25:40 +02:00
* current top-level transaction. Currently, two-phase commit does not
* persist and restore this variable. So long as all callers use
* PreventInTransactionBlock(), that omission has no consequences.
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
*/
void
ForceSyncCommit(void)
{
forceSyncCommit = true;
}
Fix very old bug which made tuples changed/inserted by a commnd visible to command itself (so we had multiple update of updated tuples, etc).
1997-08-29 11:05:25 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
* StartTransaction stuff
* ----------------------------------------------------------------
*/
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AtStart_Cache
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AtStart_Cache(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Clean up some longstanding problems in shared-cache invalidation. SI messages now include the relevant database OID, so that operations in one database do not cause useless cache flushes in backends attached to other databases. Declare SI messages properly using a union, to eliminate the former assumption that Oid is the same size as int or Index. Rewrite the nearly-unreadable code in inval.c, and document it better. Arrange for catcache flushes at end of command/transaction to happen before relcache flushes do --- this avoids loading a new tuple into the catcache while setting up new relcache entry, only to have it be flushed again immediately.
2001-06-19 21:42:16 +02:00
AcceptInvalidationMessages();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AtStart_Memory
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AtStart_Memory(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
TransactionState s = CurrentTransactionState;
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/*
* If this is the first time through, create a private context for
* AbortTransaction to work in. By reserving some space now, we can
* insulate AbortTransaction from out-of-memory scenarios. Like
* ErrorContext, we set it up with slow growth rate and a nonzero minimum
* size, so that space will be reserved immediately.
*/
if (TransactionAbortContext == NULL)
TransactionAbortContext =
Simplify use of AllocSetContextCreate() wrapper macro. We can allow this macro to accept either abbreviated or non-abbreviated allocation parameters by making use of __VA_ARGS__. As noted by Andres Freund, it's unlikely that any compiler would have __builtin_constant_p but not __VA_ARGS__, so this gives up little or no error checking, and it avoids a minor but annoying API break for extensions. With this change, there is no reason for anybody to call AllocSetContextCreateExtended directly, so in HEAD I renamed it to AllocSetContextCreateInternal. It's probably too late for an ABI break like that in 11, though. Discussion: https://postgr.es/m/20181012170355.bhxi273skjt6sag4@alap3.anarazel.de
2018-10-12 20:26:56 +02:00
AllocSetContextCreate(TopMemoryContext,
"TransactionAbortContext",
32 * 1024,
32 * 1024,
32 * 1024);
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
Portal and memory management infrastructure for extended query protocol. Both plannable queries and utility commands are now always executed within Portals, which have been revamped so that they can handle the load (they used to be good only for single SELECT queries). Restructure code to push command-completion-tag selection logic out of postgres.c, so that it won't have to be duplicated between simple and extended queries. initdb forced due to addition of a field to Query nodes.
2003-05-02 22:54:36 +02:00
* We shouldn't have a transaction context already.
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
*/
Assert(TopTransactionContext == NULL);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
* Create a toplevel context for the transaction.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
TopTransactionContext =
AllocSetContextCreate(TopMemoryContext,
"TopTransactionContext",
Add macros to make AllocSetContextCreate() calls simpler and safer. I found that half a dozen (nearly 5%) of our AllocSetContextCreate calls had typos in the context-sizing parameters. While none of these led to especially significant problems, they did create minor inefficiencies, and it's now clear that expecting people to copy-and-paste those calls accurately is not a great idea. Let's reduce the risk of future errors by introducing single macros that encapsulate the common use-cases. Three such macros are enough to cover all but two special-purpose contexts; those two calls can be left as-is, I think. While this patch doesn't in itself improve matters for third-party extensions, it doesn't break anything for them either, and they can gradually adopt the simplified notation over time. In passing, change TopMemoryContext to use the default allocation parameters. Formerly it could only be extended 8K at a time. That was probably reasonable when this code was written; but nowadays we create many more contexts than we did then, so that it's not unusual to have a couple hundred K in TopMemoryContext, even without considering various dubious code that sticks other things there. There seems no good reason not to let it use growing blocks like most other contexts. Back-patch to 9.6, mostly because that's still close enough to HEAD that it's easy to do so, and keeping the branches in sync can be expected to avoid some future back-patching pain. The bugs fixed by these changes don't seem to be significant enough to justify fixing them further back. Discussion: <21072.1472321324@sss.pgh.pa.us>
2016-08-27 23:50:38 +02:00
ALLOCSET_DEFAULT_SIZES);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* In a top-level transaction, CurTransactionContext is the same as
* TopTransactionContext.
*/
CurTransactionContext = TopTransactionContext;
s->curTransactionContext = CurTransactionContext;
/* Make the CurTransactionContext active. */
MemoryContextSwitchTo(CurTransactionContext);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
* AtStart_ResourceOwner
*/
static void
AtStart_ResourceOwner(void)
{
TransactionState s = CurrentTransactionState;
/*
* We shouldn't have a transaction resource owner already.
*/
Assert(TopTransactionResourceOwner == NULL);
/*
* Create a toplevel resource owner for the transaction.
*/
s->curTransactionOwner = ResourceOwnerCreate(NULL, "TopTransaction");
TopTransactionResourceOwner = s->curTransactionOwner;
CurTransactionResourceOwner = s->curTransactionOwner;
CurrentResourceOwner = s->curTransactionOwner;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* ----------------------------------------------------------------
* StartSubTransaction stuff
* ----------------------------------------------------------------
*/
/*
* AtSubStart_Memory
*/
static void
AtSubStart_Memory(void)
{
TransactionState s = CurrentTransactionState;
Assert(CurTransactionContext != NULL);
/*
* Create a CurTransactionContext, which will be used to hold data that
* survives subtransaction commit but disappears on subtransaction abort.
* We make it a child of the immediate parent's CurTransactionContext.
*/
CurTransactionContext = AllocSetContextCreate(CurTransactionContext,
"CurTransactionContext",
Add macros to make AllocSetContextCreate() calls simpler and safer. I found that half a dozen (nearly 5%) of our AllocSetContextCreate calls had typos in the context-sizing parameters. While none of these led to especially significant problems, they did create minor inefficiencies, and it's now clear that expecting people to copy-and-paste those calls accurately is not a great idea. Let's reduce the risk of future errors by introducing single macros that encapsulate the common use-cases. Three such macros are enough to cover all but two special-purpose contexts; those two calls can be left as-is, I think. While this patch doesn't in itself improve matters for third-party extensions, it doesn't break anything for them either, and they can gradually adopt the simplified notation over time. In passing, change TopMemoryContext to use the default allocation parameters. Formerly it could only be extended 8K at a time. That was probably reasonable when this code was written; but nowadays we create many more contexts than we did then, so that it's not unusual to have a couple hundred K in TopMemoryContext, even without considering various dubious code that sticks other things there. There seems no good reason not to let it use growing blocks like most other contexts. Back-patch to 9.6, mostly because that's still close enough to HEAD that it's easy to do so, and keeping the branches in sync can be expected to avoid some future back-patching pain. The bugs fixed by these changes don't seem to be significant enough to justify fixing them further back. Discussion: <21072.1472321324@sss.pgh.pa.us>
2016-08-27 23:50:38 +02:00
ALLOCSET_DEFAULT_SIZES);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->curTransactionContext = CurTransactionContext;
/* Make the CurTransactionContext active. */
MemoryContextSwitchTo(CurTransactionContext);
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
* AtSubStart_ResourceOwner
*/
static void
AtSubStart_ResourceOwner(void)
{
TransactionState s = CurrentTransactionState;
Assert(s->parent != NULL);
/*
* Create a resource owner for the subtransaction. We make it a child of
* the immediate parent's resource owner.
*/
s->curTransactionOwner =
ResourceOwnerCreate(s->parent->curTransactionOwner,
"SubTransaction");
CurTransactionResourceOwner = s->curTransactionOwner;
CurrentResourceOwner = s->curTransactionOwner;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
* CommitTransaction stuff
* ----------------------------------------------------------------
*/
Restructure local-buffer handling per recent pghackers discussion. The local buffer manager is no longer used for newly-created relations (unless they are TEMP); a new non-TEMP relation goes through the shared bufmgr and thus will participate normally in checkpoints. But TEMP relations use the local buffer manager throughout their lifespan. Also, operations in TEMP relations are not logged in WAL, thus improving performance. Since it's no longer necessary to fsync relations as they move out of the local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code is no longer needed and has been removed: there's no concept of a dirty relation anymore in md.c/fd.c, and we never fsync anything but WAL. Still TODO: improve local buffer management algorithms so that it would be reasonable to increase NLocBuffer.
2002-08-06 04:36:35 +02:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* RecordTransactionCommit
Don't take ProcArrayLock while exiting a transaction that has no XID; there is no need for serialization against snapshot-taking because the xact doesn't affect anyone else's snapshot anyway. Per discussion. Also, move various info about the interlocking of transactions and snapshots out of code comments and into a hopefully-more-cohesive discussion in access/transam/README. Also, remove a couple of now-obsolete comments about having to force some WAL to be written to persuade RecordTransactionCommit to do its thing.
2007-09-07 22:59:26 +02:00
*
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
* Returns latest XID among xact and its children, or InvalidTransactionId
* if the xact has no XID. (We compute that here just because it's easier.)
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
*
* If you change this function, see RecordTransactionCommitPrepared also.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Remove old-style VACUUM FULL (which was known for a little while as VACUUM FULL INPLACE), along with a boatload of subsidiary code and complexity. Per discussion, the use case for this method of vacuuming is no longer large enough to justify maintaining it; not to mention that we don't wish to invest the work that would be needed to make it play nicely with Hot Standby. Aside from the code directly related to old-style VACUUM FULL, this commit removes support for certain WAL record types that could only be generated within VACUUM FULL, redirect-pointer removal in heap_page_prune, and nontransactional generation of cache invalidation sinval messages (the last being the sticking point for Hot Standby). We still have to retain all code that copes with finding HEAP_MOVED_OFF and HEAP_MOVED_IN flag bits on existing tuples. This can't be removed as long as we want to support in-place update from pre-9.0 databases.
2010-02-08 05:33:55 +01:00
static TransactionId
RecordTransactionCommit(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
TransactionId xid = GetTopTransactionIdIfAny();
bool markXidCommitted = TransactionIdIsValid(xid);
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
TransactionId latestXid = InvalidTransactionId;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
int nrels;
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
RelFileLocator *rels;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
int nchildren;
TransactionId *children;
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
int ndroppedstats = 0;
xl_xact_stats_item *droppedstats = NULL;
Make RecordTransactionCommit() respect wal_level. Since the only purpose of WAL-loggin SharedInvalidationMessages is to support Hot Standby operation, they needn't be included when wal_level < hot_standby. Back-patch to 9.0. Review by Heikki Linnakanagas and Fujii Masao.
2010-08-13 17:42:21 +02:00
int nmsgs = 0;
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
SharedInvalidationMessage *invalMessages = NULL;
Make RecordTransactionCommit() respect wal_level. Since the only purpose of WAL-loggin SharedInvalidationMessages is to support Hot Standby operation, they needn't be included when wal_level < hot_standby. Back-patch to 9.0. Review by Heikki Linnakanagas and Fujii Masao.
2010-08-13 17:42:21 +02:00
bool RelcacheInitFileInval = false;
Allow transactions that don't write WAL to commit asynchronously. This case can arise if a transaction has written data, but only to temporary tables. Loss of the commit record in case of a crash won't matter, because the temporary tables will be lost anyway. Reviewed by Heikki Linnakangas and Simon Riggs.
2010-12-20 18:59:33 +01:00
bool wrote_xlog;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
WAL Log invalidations at command end with wal_level=logical. When wal_level=logical, write invalidations at command end into WAL so that decoding can use this information. This patch is required to allow the streaming of in-progress transactions in logical decoding.  The actual work to allow streaming will be committed as a separate patch. We still add the invalidations to the cache and write them to WAL at commit time in RecordTransactionCommit(). This uses the existing XLOG_INVALIDATIONS xlog record type, from the RM_STANDBY_ID resource manager (see LogStandbyInvalidations for details). So existing code relying on those invalidations (e.g. redo) does not need to be changed. The invalidations written at command end uses a new xlog record type XLOG_XACT_INVALIDATIONS, from RM_XACT_ID resource manager. See LogLogicalInvalidations for details. These new xlog records are ignored by existing redo procedures, which still rely on the invalidations written to commit records. The invalidations are decoded and accumulated in top-transaction, and then executed during replay.  This obviates the need to decode the invalidations as part of a commit record. Bump XLOG_PAGE_MAGIC, since this introduces XLOG_XACT_INVALIDATIONS. Author: Dilip Kumar, Tomas Vondra, Amit Kapila Reviewed-by: Amit Kapila Tested-by: Neha Sharma and Mahendra Singh Thalor Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2020-07-23 04:49:07 +02:00
/*
* Log pending invalidations for logical decoding of in-progress
* transactions. Normally for DDLs, we log this at each command end,
* however, for certain cases where we directly update the system table
* without a transaction block, the invalidations are not logged till this
* time.
*/
if (XLogLogicalInfoActive())
LogLogicalInvalidations();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* Get data needed for commit record */
Include the backend ID in the relpath of temporary relations. This allows us to reliably remove all leftover temporary relation files on cluster startup without reference to system catalogs or WAL; therefore, we no longer include temporary relations in XLOG_XACT_COMMIT and XLOG_XACT_ABORT WAL records. Since these changes require including a backend ID in each SharedInvalSmgrMsg, the size of the SharedInvalidationMessage.id field has been reduced from two bytes to one, and the maximum number of connections has been reduced from INT_MAX / 4 to 2^23-1. It would be possible to remove these restrictions by increasing the size of SharedInvalidationMessage by 4 bytes, but right now that doesn't seem like a good trade-off. Review by Jaime Casanova and Tom Lane.
2010-08-13 22:10:54 +02:00
nrels = smgrGetPendingDeletes(true, &rels);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
nchildren = xactGetCommittedChildren(&children);
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
ndroppedstats = pgstat_get_transactional_drops(true, &droppedstats);
Make RecordTransactionCommit() respect wal_level. Since the only purpose of WAL-loggin SharedInvalidationMessages is to support Hot Standby operation, they needn't be included when wal_level < hot_standby. Back-patch to 9.0. Review by Heikki Linnakanagas and Fujii Masao.
2010-08-13 17:42:21 +02:00
if (XLogStandbyInfoActive())
nmsgs = xactGetCommittedInvalidationMessages(&invalMessages,
&RelcacheInitFileInval);
Replace XLogRecPtr struct with a 64-bit integer. This simplifies code that needs to do arithmetic on XLogRecPtrs. To avoid changing on-disk format of data pages, the LSN on data pages is still stored in the old format. That should keep pg_upgrade happy. However, we have XLogRecPtrs embedded in the control file, and in the structs that are sent over the replication protocol, so this changes breaks compatibility of pg_basebackup and server. I didn't do anything about this in this patch, per discussion on -hackers, the right thing to do would to be to change the replication protocol to be architecture-independent, so that you could use a newer version of pg_receivexlog, for example, against an older server version.
2012-06-24 17:51:37 +02:00
wrote_xlog = (XactLastRecEnd != 0);
pgindent run for 9.0
2010-02-26 03:01:40 +01:00
Replace implementation of pg_log as a relation accessed through the buffer manager with 'pg_clog', a specialized access method modeled on pg_xlog. This simplifies startup (don't need to play games to open pg_log; among other things, OverrideTransactionSystem goes away), should improve performance a little, and opens the door to recycling commit log space by removing no-longer-needed segments of the commit log. Actual recycling is not there yet, but I felt I should commit this part separately since it'd still be useful if we chose not to do transaction ID wraparound.
2001-08-25 20:52:43 +02:00
/*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* If we haven't been assigned an XID yet, we neither can, nor do we want
* to write a COMMIT record.
Replace implementation of pg_log as a relation accessed through the buffer manager with 'pg_clog', a specialized access method modeled on pg_xlog. This simplifies startup (don't need to play games to open pg_log; among other things, OverrideTransactionSystem goes away), should improve performance a little, and opens the door to recycling commit log space by removing no-longer-needed segments of the commit log. Actual recycling is not there yet, but I felt I should commit this part separately since it'd still be useful if we chose not to do transaction ID wraparound.
2001-08-25 20:52:43 +02:00
*/
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
if (!markXidCommitted)
WAL misc
2000-10-24 11:56:23 +02:00
{
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
Fix inconsistencies and typos in the tree, take 10 This addresses some issues with unnecessary code comments, fixes various typos in docs and comments, and removes some orphaned structures and definitions. Author: Alexander Lakhin Discussion: https://postgr.es/m/9aabc775-5494-b372-8bcb-4dfc0bd37c68@gmail.com
2019-08-13 06:53:41 +02:00
* We expect that every RelationDropStorage is followed by a catalog
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* update, and hence XID assignment, so we shouldn't get here with any
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
* pending deletes. Same is true for dropping stats.
*
* Use a real test not just an Assert to check this, since it's a bit
* fragile.
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
*/
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (nrels != 0 || ndroppedstats != 0)
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
elog(ERROR, "cannot commit a transaction that deleted files but has no xid");
/* Can't have child XIDs either; AssignTransactionId enforces this */
Assert(nchildren == 0);
pgindent run for 8.3.
2007-11-15 22:14:46 +01:00
Emit invalidations to standby for transactions without xid. So far, when a transaction with pending invalidations, but without an assigned xid, committed, we simply ignored those invalidation messages. That's problematic, because those are actually sent for a reason. Known symptoms of this include that existing sessions on a hot-standby replica sometimes fail to notice new concurrently built indexes and visibility map updates. The solution is to WAL log such invalidations in transactions without an xid. We considered to alternatively force-assign an xid, but that'd be problematic for vacuum, which might be run in systems with few xids. Important: This adds a new WAL record, but as the patch has to be back-patched, we can't bump the WAL page magic. This means that standbys have to be updated before primaries; otherwise "PANIC: standby_redo: unknown op code 32" errors can be encountered. XXX: Reported-By: Васильев Дмитрий, Masahiko Sawada Discussion: CAB-SwXY6oH=9twBkXJtgR4UC1NqT-vpYAtxCseME62ADwyK5OA@mail.gmail.com CAD21AoDpZ6Xjg=gFrGPnSn4oTRRcwK1EBrWCq9OqOHuAcMMC=w@mail.gmail.com
2016-04-24 04:18:00 +02:00
/*
* Transactions without an assigned xid can contain invalidation
* messages (e.g. explicit relcache invalidations or catcache
* invalidations for inplace updates); standbys need to process those.
* We can't emit a commit record without an xid, and we don't want to
* force assigning an xid, because that'd be problematic for e.g.
* vacuum. Hence we emit a bespoke record for the invalidations. We
* don't want to use that in case a commit record is emitted, so they
* happen synchronously with commits (besides not wanting to emit more
Fix typo. Masahiko Sawada Discussion: http://postgr.es/m/CAD21AoCq_QG6UEo6yb-purmhLQjDLsCA2_B+8Wh3ah9P2-XmrQ@mail.gmail.com
2017-11-28 14:18:00 +01:00
* WAL records).
Emit invalidations to standby for transactions without xid. So far, when a transaction with pending invalidations, but without an assigned xid, committed, we simply ignored those invalidation messages. That's problematic, because those are actually sent for a reason. Known symptoms of this include that existing sessions on a hot-standby replica sometimes fail to notice new concurrently built indexes and visibility map updates. The solution is to WAL log such invalidations in transactions without an xid. We considered to alternatively force-assign an xid, but that'd be problematic for vacuum, which might be run in systems with few xids. Important: This adds a new WAL record, but as the patch has to be back-patched, we can't bump the WAL page magic. This means that standbys have to be updated before primaries; otherwise "PANIC: standby_redo: unknown op code 32" errors can be encountered. XXX: Reported-By: Васильев Дмитрий, Masahiko Sawada Discussion: CAB-SwXY6oH=9twBkXJtgR4UC1NqT-vpYAtxCseME62ADwyK5OA@mail.gmail.com CAD21AoDpZ6Xjg=gFrGPnSn4oTRRcwK1EBrWCq9OqOHuAcMMC=w@mail.gmail.com
2016-04-24 04:18:00 +02:00
*/
if (nmsgs != 0)
{
LogStandbyInvalidations(nmsgs, invalMessages,
RelcacheInitFileInval);
wrote_xlog = true; /* not strictly necessary */
}
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* If we didn't create XLOG entries, we're done here; otherwise we
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
* should trigger flushing those entries the same as a commit record
* would. This will primarily happen for HOT pruning and the like; we
* want these to be flushed to disk in due time.
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
*/
Allow transactions that don't write WAL to commit asynchronously. This case can arise if a transaction has written data, but only to temporary tables. Loss of the commit record in case of a crash won't matter, because the temporary tables will be lost anyway. Reviewed by Heikki Linnakangas and Simon Riggs.
2010-12-20 18:59:33 +01:00
if (!wrote_xlog)
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
goto cleanup;
}
else
{
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
bool replorigin;
/*
* Are we using the replication origins feature? Or, in other words,
* are we replaying remote actions?
*/
replorigin = (replorigin_session_origin != InvalidRepOriginId &&
replorigin_session_origin != DoNotReplicateId);
Fix failure to guarantee that a checkpoint will write out pg_clog updates for transaction commits that occurred just before the checkpoint. This is an EXTREMELY serious bug --- kudos to Satoshi Okada for creating a reproducible test case to prove its existence.
2004-08-11 06:07:16 +02:00
/*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* Mark ourselves as within our "commit critical section". This
* forces any concurrent checkpoint to wait until we've updated
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
* pg_xact. Without this, it is possible for the checkpoint to set
* REDO after the XLOG record but fail to flush the pg_xact update to
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* disk, leading to loss of the transaction commit if the system
* crashes a little later.
*
* Note: we could, but don't bother to, set this flag in
* RecordTransactionAbort. That's because loss of a transaction abort
* is noncritical; the presumption would be that it aborted, anyway.
*
Rename delayChkpt to delayChkptFlags. Before commit 412ad7a55639516f284cd0ef9757d6ae5c7abd43, delayChkpt was a Boolean. Now it's an integer. Extensions using it need to be appropriately updated, so let's rename the field to make sure that a hard compilation failure occurs. Replacing delayChkpt with delayChkptFlags made a few comments extend past 80 characters, so I reflowed them and changed some wording very slightly. The back-branches will need a different change to restore compatibility with existing minor releases; this is just for master. Per suggestion from Tom Lane. Discussion: http://postgr.es/m/a7880f4d-1d74-582a-ada7-dad168d046d1@enterprisedb.com
2022-04-08 17:44:17 +02:00
* It's safe to change the delayChkptFlags flag of our own backend
* without holding the ProcArrayLock, since we're the only one
* modifying it. This makes checkpoint's determination of which xacts
* are delaying the checkpoint a bit fuzzy, but it doesn't matter.
Fix failure to guarantee that a checkpoint will write out pg_clog updates for transaction commits that occurred just before the checkpoint. This is an EXTREMELY serious bug --- kudos to Satoshi Okada for creating a reproducible test case to prove its existence.
2004-08-11 06:07:16 +02:00
*/
Rename delayChkpt to delayChkptFlags. Before commit 412ad7a55639516f284cd0ef9757d6ae5c7abd43, delayChkpt was a Boolean. Now it's an integer. Extensions using it need to be appropriately updated, so let's rename the field to make sure that a hard compilation failure occurs. Replacing delayChkpt with delayChkptFlags made a few comments extend past 80 characters, so I reflowed them and changed some wording very slightly. The back-branches will need a different change to restore compatibility with existing minor releases; this is just for master. Per suggestion from Tom Lane. Discussion: http://postgr.es/m/a7880f4d-1d74-582a-ada7-dad168d046d1@enterprisedb.com
2022-04-08 17:44:17 +02:00
Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
START_CRIT_SECTION();
Rename delayChkpt to delayChkptFlags. Before commit 412ad7a55639516f284cd0ef9757d6ae5c7abd43, delayChkpt was a Boolean. Now it's an integer. Extensions using it need to be appropriately updated, so let's rename the field to make sure that a hard compilation failure occurs. Replacing delayChkpt with delayChkptFlags made a few comments extend past 80 characters, so I reflowed them and changed some wording very slightly. The back-branches will need a different change to restore compatibility with existing minor releases; this is just for master. Per suggestion from Tom Lane. Discussion: http://postgr.es/m/a7880f4d-1d74-582a-ada7-dad168d046d1@enterprisedb.com
2022-04-08 17:44:17 +02:00
MyProc->delayChkptFlags |= DELAY_CHKPT_START;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
Remove empty function BufmgrCommit(). This function has been a no-op for over a decade. Even if bufmgr regains a need to be called during commit, it seems unlikely that the most appropriate call points would be precisely here, so it's not doing us much good as a placeholder either. Now, removing it probably doesn't save any noticeable number of cycles --- but the main call is inside the commit critical section, and the less work done there the better. Matthias van de Meent Discussion: https://postgr.es/m/CAEze2Wi1=tLKbxZnXzcD+8fYKyKqBtivVakLQC_mYBsP4Y8qVA@mail.gmail.com
2023-03-29 15:13:57 +02:00
/*
* Insert the commit XLOG record.
*/
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
XactLogCommitRecord(GetCurrentTransactionStopTimestamp(),
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
nchildren, children, nrels, rels,
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
ndroppedstats, droppedstats,
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
nmsgs, invalMessages,
Remove dead forceSync parameter of XactLogCommitRecord(). The function has been reading global variable forceSyncCommit, mirroring the intent of the caller that passed forceSync=forceSyncCommit. The other caller, RecordTransactionCommitPrepared(), passed false. Since COMMIT PREPARED can't share a transaction with any command, it certainly doesn't share a transaction with a command that sets forceSyncCommit. Reviewed by Michael Paquier. Discussion: https://postgr.es/m/20200617032615.GC2916904@rfd.leadboat.com
2020-06-20 10:25:40 +02:00
RelcacheInitFileInval,
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
MyXactFlags,
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
InvalidTransactionId, NULL /* plain commit */ );
WAL misc
2000-10-24 11:56:23 +02:00
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
if (replorigin)
/* Move LSNs forward for this replication origin */
Fix "sesssion" typo It was introduced alongside replication origins, by commit 5aa2350426c, so backpatch to 9.5. Pointed out by Fujii Masao
2015-09-29 00:13:42 +02:00
replorigin_session_advance(replorigin_session_origin_lsn,
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
XactLastRecEnd);
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
/*
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
* Record commit timestamp. The value comes from plain commit
* timestamp if there's no replication origin; otherwise, the
* timestamp was already set in replorigin_session_origin_timestamp by
* replication.
*
* We don't need to WAL-log anything here, as the commit record
* written above already contains the data.
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
*/
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
if (!replorigin || replorigin_session_origin_timestamp == 0)
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
replorigin_session_origin_timestamp = GetCurrentTransactionStopTimestamp();
Code review for transaction commit timestamps There are three main changes here: 1. No longer cause a start failure in a standby if the feature is disabled in postgresql.conf but enabled in the master. This reverts one part of commit 4f3924d9cd43; what we keep is the ability of the standby to activate/deactivate the module (which includes creating and removing segments as appropriate) during replay of such actions in the master. 2. Replay WAL records affecting commitTS even if the feature is disabled. This means the standby will always have the same state as the master after replay. 3. Have COMMIT PREPARE record the transaction commit time as well. We were previously only applying it in the normal transaction commit path. Author: Petr Jelínek Discussion: http://www.postgresql.org/message-id/CAHGQGwHereDzzzmfxEBYcVQu3oZv6vZcgu1TPeERWbDc+gQ06g@mail.gmail.com Discussion: http://www.postgresql.org/message-id/CAHGQGwFuzfO4JscM9LCAmCDCxp_MfLvN4QdB+xWsS-FijbjTYQ@mail.gmail.com Additionally, I cleaned up nearby code related to replication origins, which I found a bit hard to follow, and fixed a couple of typos. Backpatch to 9.5, where this code was introduced. Per bug reports from Fujii Masao and subsequent discussion.
2015-09-29 19:40:56 +02:00
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
TransactionTreeSetCommitTsData(xid, nchildren, children,
Fix "sesssion" typo It was introduced alongside replication origins, by commit 5aa2350426c, so backpatch to 9.5. Pointed out by Fujii Masao
2015-09-29 00:13:42 +02:00
replorigin_session_origin_timestamp,
Remove COMMIT_TS_SETTS record. Commit 438fc4a39c prevented the WAL replay from writing COMMIT_TS_SETTS record. By this change there is no code that generates COMMIT_TS_SETTS record in PostgreSQL core. Also we can think that there are no extensions using the record because we've not received so far any complaints about the issue that commit 438fc4a39c fixed. Therefore this commit removes COMMIT_TS_SETTS record and its related code. Even without this record, the timestamp required for commit timestamp feature can be acquired from the COMMIT record. Bump WAL page magic. Reported-by: lx zou <zoulx1982@163.com> Author: Fujii Masao Reviewed-by: Alvaro Herrera Discussion: https://postgr.es/m/16931-620d0f2fdc6108f1@postgresql.org
2021-04-11 17:00:18 +02:00
replorigin_session_origin);
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
}
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
Allow transactions that don't write WAL to commit asynchronously. This case can arise if a transaction has written data, but only to temporary tables. Loss of the commit record in case of a crash won't matter, because the temporary tables will be lost anyway. Reviewed by Heikki Linnakangas and Simon Riggs.
2010-12-20 18:59:33 +01:00
* Check if we want to commit asynchronously. We can allow the XLOG flush
* to happen asynchronously if synchronous_commit=off, or if the current
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
* transaction has not performed any WAL-logged operation or didn't assign
Collection of typo fixes. Use "a" and "an" correctly, mostly in comments. Two error messages were also fixed (they were just elogs, so no translation work required). Two function comments in pg_proc.h were also fixed. Etsuro Fujita reported one of these, but I found a lot more with grep. Also fix a few other typos spotted while grepping for the a/an typos. For example, "consists out of ..." -> "consists of ...". Plus a "though"/ "through" mixup reported by Euler Taveira. Many of these typos were in old code, which would be nice to backpatch to make future backpatching easier. But much of the code was new, and I didn't feel like crafting separate patches for each branch. So no backpatching.
2015-05-20 15:18:11 +02:00
* an xid. The transaction can end up not writing any WAL, even if it has
* an xid, if it only wrote to temporary and/or unlogged tables. It can
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
* end up having written WAL without an xid if it did HOT pruning. In
* case of a crash, the loss of such a transaction will be irrelevant;
* temp tables will be lost anyway, unlogged tables will be truncated and
* HOT pruning will be done again later. (Given the foregoing, you might
Update comments in RecordTransactionCommit() to mention unlogged tables.
2011-01-03 16:29:22 +01:00
* think that it would be unnecessary to emit the XLOG record at all in
* this case, but we don't currently try to do that. It would certainly
* cause problems at least in Hot Standby mode, where the
* KnownAssignedXids machinery requires tracking every XID assignment. It
Merge wal_level "archive" and "hot_standby" into new name "replica" The distinction between "archive" and "hot_standby" existed only because at the time "hot_standby" was added, there was some uncertainty about stability. This is now a long time ago. We would like to move forward with simplifying the replication configuration, but this distinction is in the way, because a primary server cannot tell (without asking a standby or predicting the future) which one of these would be the appropriate level. Pick a new name for the combined setting to make it clearer that it covers all (non-logical) backup and replication uses. The old values are still accepted but are converted internally. Reviewed-by: Michael Paquier <michael.paquier@gmail.com> Reviewed-by: David Steele <david@pgmasters.net>
2016-03-01 02:01:54 +01:00
* might be OK to skip it only when wal_level < replica, but for now we
Update comments in RecordTransactionCommit() to mention unlogged tables.
2011-01-03 16:29:22 +01:00
* don't.)
Allow transactions that don't write WAL to commit asynchronously. This case can arise if a transaction has written data, but only to temporary tables. Loss of the commit record in case of a crash won't matter, because the temporary tables will be lost anyway. Reviewed by Heikki Linnakangas and Simon Riggs.
2010-12-20 18:59:33 +01:00
*
* However, if we're doing cleanup of any non-temp rels or committing any
* command that wanted to force sync commit, then we must flush XLOG
* immediately. (We must not allow asynchronous commit if there are any
* non-temp tables to be deleted, because we might delete the files before
* the COMMIT record is flushed to disk. We do allow asynchronous commit
* if all to-be-deleted tables are temporary though, since they are lost
* anyway if we crash.)
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
*/
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
if ((wrote_xlog && markXidCommitted &&
synchronous_commit > SYNCHRONOUS_COMMIT_OFF) ||
Merge synchronous_replication setting into synchronous_commit. This means one less thing to configure when setting up synchronous replication, and also avoids some ambiguity around what the behavior should be when the settings of these variables conflict. Fujii Masao, with additional hacking by me.
2011-04-04 22:13:01 +02:00
forceSyncCommit || nrels > 0)
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
{
XLogFlush(XactLastRecEnd);
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* Now we may update the CLOG, if we wrote a COMMIT record above
*/
if (markXidCommitted)
Rework subtransaction commit protocol for hot standby. This patch eliminates the marking of subtransactions as SUBCOMMITTED in pg_clog during their commit; instead they remain in-progress until main transaction commit. At main transaction commit, the commit protocol is atomic-by-page instead of one transaction at a time. To avoid a race condition with some subtransactions appearing committed before others in the case where they span more than one pg_clog page, we conserve the logic that marks them subcommitted before marking the parent committed. Simon Riggs with minor help from me
2008-10-20 21:18:18 +02:00
TransactionIdCommitTree(xid, nchildren, children);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
else
{
Restructure local-buffer handling per recent pghackers discussion. The local buffer manager is no longer used for newly-created relations (unless they are TEMP); a new non-TEMP relation goes through the shared bufmgr and thus will participate normally in checkpoints. But TEMP relations use the local buffer manager throughout their lifespan. Also, operations in TEMP relations are not logged in WAL, thus improving performance. Since it's no longer necessary to fsync relations as they move out of the local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code is no longer needed and has been removed: there's no concept of a dirty relation anymore in md.c/fd.c, and we never fsync anything but WAL. Still TODO: improve local buffer management algorithms so that it would be reasonable to increase NLocBuffer.
2002-08-06 04:36:35 +02:00
/*
Add C comment about why synchronous_commit=off behavior can lose committed transactions in a postmaster crash.
2010-06-29 20:44:58 +02:00
* Asynchronous commit case:
*
* This enables possible committed transaction loss in the case of a
* postmaster crash because WAL buffers are left unwritten. Ideally we
* could issue the WAL write without the fsync, but some
* wal_sync_methods do not allow separate write/fsync.
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* Report the latest async commit LSN, so that the WAL writer knows to
* flush this commit.
Restructure local-buffer handling per recent pghackers discussion. The local buffer manager is no longer used for newly-created relations (unless they are TEMP); a new non-TEMP relation goes through the shared bufmgr and thus will participate normally in checkpoints. But TEMP relations use the local buffer manager throughout their lifespan. Also, operations in TEMP relations are not logged in WAL, thus improving performance. Since it's no longer necessary to fsync relations as they move out of the local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code is no longer needed and has been removed: there's no concept of a dirty relation anymore in md.c/fd.c, and we never fsync anything but WAL. Still TODO: improve local buffer management algorithms so that it would be reasonable to increase NLocBuffer.
2002-08-06 04:36:35 +02:00
*/
Rename asyncCommitLSN to asyncXactLSN to reflect changed role in 9.0. Transaction aborts now record their LSN to avoid corner case behaviour in SR/HS, hence change of name of variables and functions. As pointed out by Fujii Masao. Cosmetic changes only.
2010-07-30 00:27:27 +02:00
XLogSetAsyncXactLSN(XactLastRecEnd);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* We must not immediately update the CLOG, since we didn't flush the
* XLOG. Instead, we store the LSN up to which the XLOG must be
* flushed before the CLOG may be updated.
*/
if (markXidCommitted)
Rework subtransaction commit protocol for hot standby. This patch eliminates the marking of subtransactions as SUBCOMMITTED in pg_clog during their commit; instead they remain in-progress until main transaction commit. At main transaction commit, the commit protocol is atomic-by-page instead of one transaction at a time. To avoid a race condition with some subtransactions appearing committed before others in the case where they span more than one pg_clog page, we conserve the logic that marks them subcommitted before marking the parent committed. Simon Riggs with minor help from me
2008-10-20 21:18:18 +02:00
TransactionIdAsyncCommitTree(xid, nchildren, children, XactLastRecEnd);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
WAL misc
2000-10-24 11:56:23 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* If we entered a commit critical section, leave it now, and let
* checkpoints proceed.
*/
if (markXidCommitted)
{
Rename delayChkpt to delayChkptFlags. Before commit 412ad7a55639516f284cd0ef9757d6ae5c7abd43, delayChkpt was a Boolean. Now it's an integer. Extensions using it need to be appropriately updated, so let's rename the field to make sure that a hard compilation failure occurs. Replacing delayChkpt with delayChkptFlags made a few comments extend past 80 characters, so I reflowed them and changed some wording very slightly. The back-branches will need a different change to restore compatibility with existing minor releases; this is just for master. Per suggestion from Tom Lane. Discussion: http://postgr.es/m/a7880f4d-1d74-582a-ada7-dad168d046d1@enterprisedb.com
2022-04-08 17:44:17 +02:00
MyProc->delayChkptFlags &= ~DELAY_CHKPT_START;
Add more critical-section calls: all code sections that hold spinlocks are now critical sections, so as to ensure die() won't interrupt us while we are munging shared-memory data structures. Avoid insecure intermediate states in some code that proc_exit will call, like palloc/pfree. Rename START/END_CRIT_CODE to START/END_CRIT_SECTION, since that seems to be what people tend to call them anyway, and make them be called with () like a function call, in hopes of not confusing pg_indent. I doubt that this is sufficient to make SIGTERM safe anywhere; there's just too much code that could get invoked during proc_exit().
2001-01-12 22:54:01 +01:00
END_CRIT_SECTION();
WAL misc
2000-10-24 11:56:23 +02:00
}
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
/* Compute latestXid while we have the child XIDs handy */
latestXid = TransactionIdLatest(xid, nchildren, children);
Efficient transaction-controlled synchronous replication. If a standby is broadcasting reply messages and we have named one or more standbys in synchronous_standby_names then allow users who set synchronous_replication to wait for commit, which then provides strict data integrity guarantees. Design avoids sending and receiving transaction state information so minimises bookkeeping overheads. We synchronize with the highest priority standby that is connected and ready to synchronize. Other standbys can be defined to takeover in case of standby failure. This version has very strict behaviour; more relaxed options may be added at a later date. Simon Riggs and Fujii Masao, with reviews by Yeb Havinga, Jaime Casanova, Heikki Linnakangas and Robert Haas, plus the assistance of many other design reviewers.
2011-03-06 23:49:16 +01:00
/*
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
* Wait for synchronous replication, if required. Similar to the decision
* above about using committing asynchronously we only want to wait if
Collection of typo fixes. Use "a" and "an" correctly, mostly in comments. Two error messages were also fixed (they were just elogs, so no translation work required). Two function comments in pg_proc.h were also fixed. Etsuro Fujita reported one of these, but I found a lot more with grep. Also fix a few other typos spotted while grepping for the a/an typos. For example, "consists out of ..." -> "consists of ...". Plus a "though"/ "through" mixup reported by Euler Taveira. Many of these typos were in old code, which would be nice to backpatch to make future backpatching easier. But much of the code was new, and I didn't feel like crafting separate patches for each branch. So no backpatching.
2015-05-20 15:18:11 +02:00
* this backend assigned an xid and wrote WAL. No need to wait if an xid
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
* was assigned due to temporary/unlogged tables or due to HOT pruning.
Efficient transaction-controlled synchronous replication. If a standby is broadcasting reply messages and we have named one or more standbys in synchronous_standby_names then allow users who set synchronous_replication to wait for commit, which then provides strict data integrity guarantees. Design avoids sending and receiving transaction state information so minimises bookkeeping overheads. We synchronize with the highest priority standby that is connected and ready to synchronize. Other standbys can be defined to takeover in case of standby failure. This version has very strict behaviour; more relaxed options may be added at a later date. Simon Riggs and Fujii Masao, with reviews by Yeb Havinga, Jaime Casanova, Heikki Linnakangas and Robert Haas, plus the assistance of many other design reviewers.
2011-03-06 23:49:16 +01:00
*
* Note that at this stage we have marked clog, but still show as running
* in the procarray and continue to hold locks.
*/
Reconsider when to wait for WAL flushes/syncrep during commit. Up to now RecordTransactionCommit() waited for WAL to be flushed (if synchronous_commit != off) and to be synchronously replicated (if enabled), even if a transaction did not have a xid assigned. The primary reason for that is that sequence's nextval() did not assign a xid, but are worthwhile to wait for on commit. This can be problematic because sometimes read only transactions do write WAL, e.g. HOT page prune records. That then could lead to read only transactions having to wait during commit. Not something people expect in a read only transaction. This lead to such strange symptoms as backends being seemingly stuck during connection establishment when all synchronous replicas are down. Especially annoying when said stuck connection is the standby trying to reconnect to allow syncrep again... This behavior also is involved in a rather complicated <= 9.4 bug where the transaction started by catchup interrupt processing waited for syncrep using latches, but didn't get the wakeup because it was already running inside the same overloaded signal handler. Fix the issue here doesn't properly solve that issue, merely papers over the problems. In 9.5 catchup interrupts aren't processed out of signal handlers anymore. To fix all this, make nextval() acquire a top level xid, and only wait for transaction commit if a transaction both acquired a xid and emitted WAL records. If only a xid has been assigned we don't uselessly want to wait just because of writes to temporary/unlogged tables; if only WAL has been written we don't want to wait just because of HOT prunes. The xid assignment in nextval() is unlikely to cause overhead in real-world workloads. For one it only happens SEQ_LOG_VALS/32 values anyway, for another only usage of nextval() without using the result in an insert or similar is affected. Discussion: 20150223165359.GF30784@awork2.anarazel.de, 369698E947874884A77849D8FE3680C2@maumau, 5CF4ABBA67674088B3941894E22A0D25@maumau Per complaint from maumau and Thom Brown Backpatch all the way back; 9.0 doesn't have syncrep, but it seems better to be consistent behavior across all maintained branches.
2015-02-26 12:50:07 +01:00
if (wrote_xlog && markXidCommitted)
Add new replication mode synchronous_commit = 'remote_apply'. In this mode, the master waits for the transaction to be applied on the remote side, not just written to disk. That means that you can count on a transaction started on the standby to see all commits previously acknowledged by the master. To make this work, the standby sends a reply after replaying each commit record generated with synchronous_commit >= 'remote_apply'. This introduces a small inefficiency: the extra replies will be sent even by standbys that aren't the current synchronous standby. But previously-existing synchronous_commit levels make no attempt at all to optimize which replies are sent based on what the primary cares about, so this is no worse, and at least avoids any extra replies for people not using the feature at all. Thomas Munro, reviewed by Michael Paquier and by me. Some additional tweaks by me.
2016-03-30 03:16:12 +02:00
SyncRepWaitForLSN(XactLastRecEnd, true);
Efficient transaction-controlled synchronous replication. If a standby is broadcasting reply messages and we have named one or more standbys in synchronous_standby_names then allow users who set synchronous_replication to wait for commit, which then provides strict data integrity guarantees. Design avoids sending and receiving transaction state information so minimises bookkeeping overheads. We synchronize with the highest priority standby that is connected and ready to synchronize. Other standbys can be defined to takeover in case of standby failure. This version has very strict behaviour; more relaxed options may be added at a later date. Simon Riggs and Fujii Masao, with reviews by Yeb Havinga, Jaime Casanova, Heikki Linnakangas and Robert Haas, plus the assistance of many other design reviewers.
2011-03-06 23:49:16 +01:00
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
/* remember end of last commit record */
XactLastCommitEnd = XactLastRecEnd;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Reset XactLastRecEnd until the next transaction writes something */
Replace XLogRecPtr struct with a 64-bit integer. This simplifies code that needs to do arithmetic on XLogRecPtrs. To avoid changing on-disk format of data pages, the LSN on data pages is still stored in the old format. That should keep pg_upgrade happy. However, we have XLogRecPtrs embedded in the control file, and in the structs that are sent over the replication protocol, so this changes breaks compatibility of pg_basebackup and server. I didn't do anything about this in this patch, per discussion on -hackers, the right thing to do would to be to change the replication protocol to be architecture-independent, so that you could use a newer version of pg_receivexlog, for example, against an older server version.
2012-06-24 17:51:37 +02:00
XactLastRecEnd = 0;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
cleanup:
/* Clean up local data */
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
if (rels)
pfree(rels);
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (ndroppedstats)
pfree(droppedstats);
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
return latestXid;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
* AtCCI_LocalCache
Improve cache invalidation handling. Eespecially this would fix TODO * elog() flushes cache, try invalidating just entries from current xact, perhaps using invalidation cache
2000-01-10 07:30:56 +01:00
*/
static void
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
AtCCI_LocalCache(void)
Improve cache invalidation handling. Eespecially this would fix TODO * elog() flushes cache, try invalidating just entries from current xact, perhaps using invalidation cache
2000-01-10 07:30:56 +01:00
{
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
/*
* Make any pending relation map changes visible. We must do this before
* processing local sinval messages, so that the map changes will get
* reflected into the relcache when relcache invals are processed.
*/
AtCCI_RelationMap();
Improve cache invalidation handling. Eespecially this would fix TODO * elog() flushes cache, try invalidating just entries from current xact, perhaps using invalidation cache
2000-01-10 07:30:56 +01:00
/*
* Make catalog changes visible to me for the next command.
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CommandEndInvalidationMessages();
Improve cache invalidation handling. Eespecially this would fix TODO * elog() flushes cache, try invalidating just entries from current xact, perhaps using invalidation cache
2000-01-10 07:30:56 +01:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AtCommit_Memory
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AtCommit_Memory(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
/*
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
* Now that we're "out" of a transaction, have the system allocate things
* in the top memory context instead of per-transaction contexts.
Release allocated memory during AtAbort_Memory.
1999-05-13 02:34:57 +02:00
*/
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
MemoryContextSwitchTo(TopMemoryContext);
Release allocated memory during AtAbort_Memory.
1999-05-13 02:34:57 +02:00
/*
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
* Release all transaction-local memory.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
New memmgr logic in xact.c failed if AbortTransaction() is called when there is no open transaction.
2000-07-02 04:28:38 +02:00
Assert(TopTransactionContext != NULL);
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
MemoryContextDelete(TopTransactionContext);
TopTransactionContext = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CurTransactionContext = NULL;
CurrentTransactionState->curTransactionContext = NULL;
}
/* ----------------------------------------------------------------
* CommitSubTransaction stuff
* ----------------------------------------------------------------
*/
/*
* AtSubCommit_Memory
*/
static void
AtSubCommit_Memory(void)
{
TransactionState s = CurrentTransactionState;
Assert(s->parent != NULL);
/* Return to parent transaction level's memory context. */
CurTransactionContext = s->parent->curTransactionContext;
MemoryContextSwitchTo(CurTransactionContext);
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
/*
* Ordinarily we cannot throw away the child's CurTransactionContext,
* since the data it contains will be needed at upper commit. However, if
* there isn't actually anything in it, we can throw it away. This avoids
* a small memory leak in the common case of "trivial" subxacts.
*/
if (MemoryContextIsEmpty(s->curTransactionContext))
{
MemoryContextDelete(s->curTransactionContext);
s->curTransactionContext = NULL;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
/*
* AtSubCommit_childXids
*
* Pass my own XID and my child XIDs up to my parent as committed children.
*/
static void
AtSubCommit_childXids(void)
{
TransactionState s = CurrentTransactionState;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
int new_nChildXids;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Assert(s->parent != NULL);
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
/*
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
* The parent childXids array will need to hold my XID and all my
* childXids, in addition to the XIDs already there.
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
*/
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
new_nChildXids = s->parent->nChildXids + s->nChildXids + 1;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
/* Allocate or enlarge the parent array if necessary */
if (s->parent->maxChildXids < new_nChildXids)
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
{
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
int new_maxChildXids;
TransactionId *new_childXids;
Standard pgindent run for 8.1.
2005-10-15 04:49:52 +02:00
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
/*
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
* Make it 2x what's needed right now, to avoid having to enlarge it
* repeatedly. But we can't go above MaxAllocSize. (The latter limit
* is what ensures that we don't need to worry about integer overflow
* here or in the calculation of new_nChildXids.)
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
*/
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
new_maxChildXids = Min(new_nChildXids * 2,
(int) (MaxAllocSize / sizeof(TransactionId)));
if (new_maxChildXids < new_nChildXids)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("maximum number of committed subtransactions (%d) exceeded",
(int) (MaxAllocSize / sizeof(TransactionId)))));
/*
* We keep the child-XID arrays in TopTransactionContext; this avoids
* setting up child-transaction contexts for what might be just a few
* bytes of grandchild XIDs.
*/
if (s->parent->childXids == NULL)
new_childXids =
MemoryContextAlloc(TopTransactionContext,
new_maxChildXids * sizeof(TransactionId));
else
new_childXids = repalloc(s->parent->childXids,
new_maxChildXids * sizeof(TransactionId));
s->parent->childXids = new_childXids;
s->parent->maxChildXids = new_maxChildXids;
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
/*
* Copy all my XIDs to parent's array.
*
* Note: We rely on the fact that the XID of a child always follows that
* of its parent. By copying the XID of this subtransaction before the
* XIDs of its children, we ensure that the array stays ordered. Likewise,
* all XIDs already in the array belong to subtransactions started and
* subcommitted before us, so their XIDs must precede ours.
*/
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->parent->childXids[s->parent->nChildXids] = XidFromFullTransactionId(s->fullTransactionId);
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
if (s->nChildXids > 0)
memcpy(&s->parent->childXids[s->parent->nChildXids + 1],
s->childXids,
s->nChildXids * sizeof(TransactionId));
s->parent->nChildXids = new_nChildXids;
/* Release child's array to avoid leakage */
if (s->childXids != NULL)
pfree(s->childXids);
/* We must reset these to avoid double-free if fail later in commit */
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
* AbortTransaction stuff
* ----------------------------------------------------------------
*/
Restructure local-buffer handling per recent pghackers discussion. The local buffer manager is no longer used for newly-created relations (unless they are TEMP); a new non-TEMP relation goes through the shared bufmgr and thus will participate normally in checkpoints. But TEMP relations use the local buffer manager throughout their lifespan. Also, operations in TEMP relations are not logged in WAL, thus improving performance. Since it's no longer necessary to fsync relations as they move out of the local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code is no longer needed and has been removed: there's no concept of a dirty relation anymore in md.c/fd.c, and we never fsync anything but WAL. Still TODO: improve local buffer management algorithms so that it would be reasonable to increase NLocBuffer.
2002-08-06 04:36:35 +02:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* RecordTransactionAbort
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
*
* Returns latest XID among xact and its children, or InvalidTransactionId
* if the xact has no XID. (We compute that here just because it's easier.)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
static TransactionId
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
RecordTransactionAbort(bool isSubXact)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
TransactionId xid = GetCurrentTransactionIdIfAny();
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
TransactionId latestXid;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
int nrels;
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
RelFileLocator *rels;
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
int ndroppedstats = 0;
xl_xact_stats_item *droppedstats = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
int nchildren;
TransactionId *children;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TimestampTz xact_time;
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
bool replorigin;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
XLOG (and related) changes: * Store two past checkpoint locations, not just one, in pg_control. On startup, we fall back to the older checkpoint if the newer one is unreadable. Also, a physical copy of the newest checkpoint record is kept in pg_control for possible use in disaster recovery (ie, complete loss of pg_xlog). Also add a version number for pg_control itself. Remove archdir from pg_control; it ought to be a GUC parameter, not a special case (not that it's implemented yet anyway). * Suppress successive checkpoint records when nothing has been entered in the WAL log since the last one. This is not so much to avoid I/O as to make it actually useful to keep track of the last two checkpoints. If the things are right next to each other then there's not a lot of redundancy gained... * Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs on alternate bytes. Polynomial borrowed from ECMA DLT1 standard. * Fix XLOG record length handling so that it will work at BLCKSZ = 32k. * Change XID allocation to work more like OID allocation. (This is of dubious necessity, but I think it's a good idea anyway.) * Fix a number of minor bugs, such as off-by-one logic for XLOG file wraparound at the 4 gig mark. * Add documentation and clean up some coding infelicities; move file format declarations out to include files where planned contrib utilities can get at them. * Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also possible to force a checkpoint by sending SIGUSR1 to the postmaster (undocumented feature...) * Defend against kill -9 postmaster by storing shmem block's key and ID in postmaster.pid lockfile, and checking at startup to ensure that no processes are still connected to old shmem block (if it still exists). * Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency stop, for symmetry with postmaster and xlog utilities. Clean up signal handling in bootstrap.c so that xlog utilities launched by postmaster will react to signals better. * Standalone bootstrap now grabs lockfile in target directory, as added insurance against running it in parallel with live postmaster.
2001-03-13 02:17:06 +01:00
/*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* If we haven't been assigned an XID, nobody will care whether we aborted
* or not. Hence, we're done in that case. It does not matter if we have
* rels to delete (note that this routine is not responsible for actually
* deleting 'em). We cannot have any child XIDs, either.
XLOG (and related) changes: * Store two past checkpoint locations, not just one, in pg_control. On startup, we fall back to the older checkpoint if the newer one is unreadable. Also, a physical copy of the newest checkpoint record is kept in pg_control for possible use in disaster recovery (ie, complete loss of pg_xlog). Also add a version number for pg_control itself. Remove archdir from pg_control; it ought to be a GUC parameter, not a special case (not that it's implemented yet anyway). * Suppress successive checkpoint records when nothing has been entered in the WAL log since the last one. This is not so much to avoid I/O as to make it actually useful to keep track of the last two checkpoints. If the things are right next to each other then there's not a lot of redundancy gained... * Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs on alternate bytes. Polynomial borrowed from ECMA DLT1 standard. * Fix XLOG record length handling so that it will work at BLCKSZ = 32k. * Change XID allocation to work more like OID allocation. (This is of dubious necessity, but I think it's a good idea anyway.) * Fix a number of minor bugs, such as off-by-one logic for XLOG file wraparound at the 4 gig mark. * Add documentation and clean up some coding infelicities; move file format declarations out to include files where planned contrib utilities can get at them. * Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also possible to force a checkpoint by sending SIGUSR1 to the postmaster (undocumented feature...) * Defend against kill -9 postmaster by storing shmem block's key and ID in postmaster.pid lockfile, and checking at startup to ensure that no processes are still connected to old shmem block (if it still exists). * Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency stop, for symmetry with postmaster and xlog utilities. Clean up signal handling in bootstrap.c so that xlog utilities launched by postmaster will react to signals better. * Standalone bootstrap now grabs lockfile in target directory, as added insurance against running it in parallel with live postmaster.
2001-03-13 02:17:06 +01:00
*/
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
if (!TransactionIdIsValid(xid))
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
{
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Reset XactLastRecEnd until the next transaction writes something */
if (!isSubXact)
Replace XLogRecPtr struct with a 64-bit integer. This simplifies code that needs to do arithmetic on XLogRecPtrs. To avoid changing on-disk format of data pages, the LSN on data pages is still stored in the old format. That should keep pg_upgrade happy. However, we have XLogRecPtrs embedded in the control file, and in the structs that are sent over the replication protocol, so this changes breaks compatibility of pg_basebackup and server. I didn't do anything about this in this patch, per discussion on -hackers, the right thing to do would to be to change the replication protocol to be architecture-independent, so that you could use a newer version of pg_receivexlog, for example, against an older server version.
2012-06-24 17:51:37 +02:00
XactLastRecEnd = 0;
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
return InvalidTransactionId;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* We have a valid XID, so we should write an ABORT record for it.
*
* We do not flush XLOG to disk here, since the default assumption after a
* crash would be that we aborted, anyway. For the same reason, we don't
* need to worry about interlocking against checkpoint start.
*/
New WAL version - CRC and data blocks backup.
2000-12-28 14:00:29 +01:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* Check that we haven't aborted halfway through RecordTransactionCommit.
*/
if (TransactionIdDidCommit(xid))
elog(PANIC, "cannot abort transaction %u, it was already committed",
xid);
XLOG (and related) changes: * Store two past checkpoint locations, not just one, in pg_control. On startup, we fall back to the older checkpoint if the newer one is unreadable. Also, a physical copy of the newest checkpoint record is kept in pg_control for possible use in disaster recovery (ie, complete loss of pg_xlog). Also add a version number for pg_control itself. Remove archdir from pg_control; it ought to be a GUC parameter, not a special case (not that it's implemented yet anyway). * Suppress successive checkpoint records when nothing has been entered in the WAL log since the last one. This is not so much to avoid I/O as to make it actually useful to keep track of the last two checkpoints. If the things are right next to each other then there's not a lot of redundancy gained... * Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs on alternate bytes. Polynomial borrowed from ECMA DLT1 standard. * Fix XLOG record length handling so that it will work at BLCKSZ = 32k. * Change XID allocation to work more like OID allocation. (This is of dubious necessity, but I think it's a good idea anyway.) * Fix a number of minor bugs, such as off-by-one logic for XLOG file wraparound at the 4 gig mark. * Add documentation and clean up some coding infelicities; move file format declarations out to include files where planned contrib utilities can get at them. * Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also possible to force a checkpoint by sending SIGUSR1 to the postmaster (undocumented feature...) * Defend against kill -9 postmaster by storing shmem block's key and ID in postmaster.pid lockfile, and checking at startup to ensure that no processes are still connected to old shmem block (if it still exists). * Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency stop, for symmetry with postmaster and xlog utilities. Clean up signal handling in bootstrap.c so that xlog utilities launched by postmaster will react to signals better. * Standalone bootstrap now grabs lockfile in target directory, as added insurance against running it in parallel with live postmaster.
2001-03-13 02:17:06 +01:00
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
/*
* Are we using the replication origins feature? Or, in other words, are
* we replaying remote actions?
*/
replorigin = (replorigin_session_origin != InvalidRepOriginId &&
replorigin_session_origin != DoNotReplicateId);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Fetch the data we need for the abort record */
Include the backend ID in the relpath of temporary relations. This allows us to reliably remove all leftover temporary relation files on cluster startup without reference to system catalogs or WAL; therefore, we no longer include temporary relations in XLOG_XACT_COMMIT and XLOG_XACT_ABORT WAL records. Since these changes require including a backend ID in each SharedInvalSmgrMsg, the size of the SharedInvalidationMessage.id field has been reduced from two bytes to one, and the maximum number of connections has been reduced from INT_MAX / 4 to 2^23-1. It would be possible to remove these restrictions by increasing the size of SharedInvalidationMessage by 4 bytes, but right now that doesn't seem like a good trade-off. Review by Jaime Casanova and Tom Lane.
2010-08-13 22:10:54 +02:00
nrels = smgrGetPendingDeletes(false, &rels);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
nchildren = xactGetCommittedChildren(&children);
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
ndroppedstats = pgstat_get_transactional_drops(false, &droppedstats);
Restructure local-buffer handling per recent pghackers discussion. The local buffer manager is no longer used for newly-created relations (unless they are TEMP); a new non-TEMP relation goes through the shared bufmgr and thus will participate normally in checkpoints. But TEMP relations use the local buffer manager throughout their lifespan. Also, operations in TEMP relations are not logged in WAL, thus improving performance. Since it's no longer necessary to fsync relations as they move out of the local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code is no longer needed and has been removed: there's no concept of a dirty relation anymore in md.c/fd.c, and we never fsync anything but WAL. Still TODO: improve local buffer management algorithms so that it would be reasonable to increase NLocBuffer.
2002-08-06 04:36:35 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* XXX do we really need a critical section here? */
START_CRIT_SECTION();
Commit the reasonably uncontroversial parts of J.R. Nield's PITR patch, to wit: Add a header record to each WAL segment file so that it can be reliably identified. Avoid splitting WAL records across segment files (this is not strictly necessary, but makes it simpler to incorporate the header records). Make WAL entries for file creation, deletion, and truncation (as foreseen but never implemented by Vadim). Also, add support for making XLOG_SEG_SIZE configurable at compile time, similarly to BLCKSZ. Fix a couple bugs I introduced in WAL replay during recent smgr API changes. initdb is forced due to changes in pg_control contents.
2004-02-11 23:55:26 +01:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Write the ABORT record */
if (isSubXact)
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xact_time = GetCurrentTimestamp();
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
else
{
Have GetCurrentTransactionStopTimestamp() set xactStopTimestamp if unset Previously GetCurrentTransactionStopTimestamp() computed a new timestamp whenever xactStopTimestamp was unset and xactStopTimestamp was only set when a commit or abort record was written. An upcoming patch will add additional calls to GetCurrentTransactionStopTimestamp() from pgstats. To avoid computing timestamps multiple times, set xactStopTimestamp in GetCurrentTransactionStopTimestamp() if not already set. Author: Dave Page <dpage@pgadmin.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Vik Fearing <vik@postgresfriends.org> Discussion: https://postgr.es/m/20220906155325.an3xesq5o3fq36gt%40awork3.anarazel.de
2022-10-14 20:11:33 +02:00
xact_time = GetCurrentTransactionStopTimestamp();
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
}
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
XactLogAbortRecord(xact_time,
nchildren, children,
nrels, rels,
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
ndroppedstats, droppedstats,
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
MyXactFlags, InvalidTransactionId,
NULL);
XLOG (and related) changes: * Store two past checkpoint locations, not just one, in pg_control. On startup, we fall back to the older checkpoint if the newer one is unreadable. Also, a physical copy of the newest checkpoint record is kept in pg_control for possible use in disaster recovery (ie, complete loss of pg_xlog). Also add a version number for pg_control itself. Remove archdir from pg_control; it ought to be a GUC parameter, not a special case (not that it's implemented yet anyway). * Suppress successive checkpoint records when nothing has been entered in the WAL log since the last one. This is not so much to avoid I/O as to make it actually useful to keep track of the last two checkpoints. If the things are right next to each other then there's not a lot of redundancy gained... * Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs on alternate bytes. Polynomial borrowed from ECMA DLT1 standard. * Fix XLOG record length handling so that it will work at BLCKSZ = 32k. * Change XID allocation to work more like OID allocation. (This is of dubious necessity, but I think it's a good idea anyway.) * Fix a number of minor bugs, such as off-by-one logic for XLOG file wraparound at the 4 gig mark. * Add documentation and clean up some coding infelicities; move file format declarations out to include files where planned contrib utilities can get at them. * Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or every CHECKPOINT_TIMEOUT seconds, whichever comes first. It is also possible to force a checkpoint by sending SIGUSR1 to the postmaster (undocumented feature...) * Defend against kill -9 postmaster by storing shmem block's key and ID in postmaster.pid lockfile, and checking at startup to ensure that no processes are still connected to old shmem block (if it still exists). * Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency stop, for symmetry with postmaster and xlog utilities. Clean up signal handling in bootstrap.c so that xlog utilities launched by postmaster will react to signals better. * Standalone bootstrap now grabs lockfile in target directory, as added insurance against running it in parallel with live postmaster.
2001-03-13 02:17:06 +01:00
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
if (replorigin)
/* Move LSNs forward for this replication origin */
replorigin_session_advance(replorigin_session_origin_lsn,
XactLastRecEnd);
Ensure that top level aborts call XLogSetAsyncCommit(). Not doing so simply leads to data waiting in wal_buffers which then causes later commits to potentially do emergency writes and for all forms of replication to be potentially delayed without need or benefit. Issue pointed out exactly by Fujii Masao, following bug report by Robert Haas on a separate though related topic.
2010-05-13 13:39:30 +02:00
/*
* Report the latest async abort LSN, so that the WAL writer knows to
* flush this abort. There's nothing to be gained by delaying this, since
* WALWriter may as well do this when it can. This is important with
* streaming replication because if we don't flush WAL regularly we will
* find that large aborts leave us with a long backlog for when commits
* occur after the abort, increasing our window of data loss should
* problems occur at that point.
*/
if (!isSubXact)
Rename asyncCommitLSN to asyncXactLSN to reflect changed role in 9.0. Transaction aborts now record their LSN to avoid corner case behaviour in SR/HS, hence change of name of variables and functions. As pointed out by Fujii Masao. Cosmetic changes only.
2010-07-30 00:27:27 +02:00
XLogSetAsyncXactLSN(XactLastRecEnd);
Ensure that top level aborts call XLogSetAsyncCommit(). Not doing so simply leads to data waiting in wal_buffers which then causes later commits to potentially do emergency writes and for all forms of replication to be potentially delayed without need or benefit. Issue pointed out exactly by Fujii Masao, following bug report by Robert Haas on a separate though related topic.
2010-05-13 13:39:30 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* Mark the transaction aborted in clog. This is not absolutely necessary
* but we may as well do it while we are here; also, in the subxact case
* it is helpful because XactLockTableWait makes use of it to avoid
* waiting for already-aborted subtransactions. It is OK to do it without
* having flushed the ABORT record to disk, because in event of a crash
* we'd be assumed to have aborted anyway.
*/
Rework subtransaction commit protocol for hot standby. This patch eliminates the marking of subtransactions as SUBCOMMITTED in pg_clog during their commit; instead they remain in-progress until main transaction commit. At main transaction commit, the commit protocol is atomic-by-page instead of one transaction at a time. To avoid a race condition with some subtransactions appearing committed before others in the case where they span more than one pg_clog page, we conserve the logic that marks them subcommitted before marking the parent committed. Simon Riggs with minor help from me
2008-10-20 21:18:18 +02:00
TransactionIdAbortTree(xid, nchildren, children);
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
END_CRIT_SECTION();
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
/* Compute latestXid while we have the child XIDs handy */
latestXid = TransactionIdLatest(xid, nchildren, children);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* If we're aborting a subtransaction, we can immediately remove failed
* XIDs from PGPROC's cache of running child XIDs. We do that here for
* subxacts, because we already have the child XID array at hand. For
* main xacts, the equivalent happens just after this function returns.
*/
if (isSubXact)
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
XidCacheRemoveRunningXids(xid, nchildren, children, latestXid);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Reset XactLastRecEnd until the next transaction writes something */
if (!isSubXact)
Replace XLogRecPtr struct with a 64-bit integer. This simplifies code that needs to do arithmetic on XLogRecPtrs. To avoid changing on-disk format of data pages, the LSN on data pages is still stored in the old format. That should keep pg_upgrade happy. However, we have XLogRecPtrs embedded in the control file, and in the structs that are sent over the replication protocol, so this changes breaks compatibility of pg_basebackup and server. I didn't do anything about this in this patch, per discussion on -hackers, the right thing to do would to be to change the replication protocol to be architecture-independent, so that you could use a newer version of pg_receivexlog, for example, against an older server version.
2012-06-24 17:51:37 +02:00
XactLastRecEnd = 0;
Repair two problems with WAL logging of sequence nextvalI() ops, as per recent pghackers discussion: force a new WAL record at first nextval after a checkpoint, and ensure that xlog is flushed to disk if a nextval record is the only thing emitted by a transaction.
2002-03-15 20:20:36 +01:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* And clean up local data */
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
if (rels)
pfree(rels);
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (ndroppedstats)
pfree(droppedstats);
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
return latestXid;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AtAbort_Memory
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AtAbort_Memory(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
* Switch into TransactionAbortContext, which should have some free space
* even if nothing else does. We'll work in this context until we've
* finished cleaning up.
*
* It is barely possible to get here when we've not been able to create
* TransactionAbortContext yet; if so use TopMemoryContext.
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
*/
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
if (TransactionAbortContext != NULL)
MemoryContextSwitchTo(TransactionAbortContext);
New memmgr logic in xact.c failed if AbortTransaction() is called when there is no open transaction.
2000-07-02 04:28:38 +02:00
else
MemoryContextSwitchTo(TopMemoryContext);
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* AtSubAbort_Memory
*/
static void
AtSubAbort_Memory(void)
{
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
Assert(TransactionAbortContext != NULL);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
MemoryContextSwitchTo(TransactionAbortContext);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
PortalRun must guard against the possibility that the portal it's running contains VACUUM or a similar command that will internally start and commit transactions. In such a case, the original caller values of CurrentMemoryContext and CurrentResourceOwner will point to objects that will be destroyed by the internal commit. We must restore these pointers to point to the newly-manufactured transaction context and resource owner, rather than possibly pointing to deleted memory. Also tweak xact.c so that AbortTransaction and AbortSubTransaction forcibly restore a sane value for CurrentResourceOwner, much as they have always done for CurrentMemoryContext. I'm not certain this is necessary but I'm feeling paranoid today. Responds to Sean Chittenden's bug report of 4-Oct.
2004-10-04 23:52:15 +02:00
/*
* AtAbort_ResourceOwner
*/
static void
AtAbort_ResourceOwner(void)
{
/*
* Make sure we have a valid ResourceOwner, if possible (else it will be
* NULL, which is OK)
*/
CurrentResourceOwner = TopTransactionResourceOwner;
}
/*
* AtSubAbort_ResourceOwner
*/
static void
AtSubAbort_ResourceOwner(void)
{
TransactionState s = CurrentTransactionState;
/* Make sure we have a valid ResourceOwner */
CurrentResourceOwner = s->curTransactionOwner;
}
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
/*
* AtSubAbort_childXids
*/
static void
AtSubAbort_childXids(void)
{
TransactionState s = CurrentTransactionState;
/*
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
* We keep the child-XID arrays in TopTransactionContext (see
* AtSubCommit_childXids). This means we'd better free the array
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
* explicitly at abort to avoid leakage.
*/
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
if (s->childXids != NULL)
pfree(s->childXids);
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* We could prune the unreportedXids array here. But we don't bother. That
* would potentially reduce number of XLOG_XACT_ASSIGNMENT records but it
* would likely introduce more CPU time into the more common paths, so we
* choose not to do that.
*/
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
}
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/* ----------------------------------------------------------------
* CleanupTransaction stuff
* ----------------------------------------------------------------
*/
Release allocated memory during AtAbort_Memory.
1999-05-13 02:34:57 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AtCleanup_Memory
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
*/
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AtCleanup_Memory(void)
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
{
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
Assert(CurrentTransactionState->parent == NULL);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Release allocated memory during AtAbort_Memory.
1999-05-13 02:34:57 +02:00
* Now that we're "out" of a transaction, have the system allocate things
* in the top memory context instead of per-transaction contexts.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Remove more (void) and fix -Wall warnings.
1997-08-13 00:55:25 +02:00
MemoryContextSwitchTo(TopMemoryContext);
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/*
* Clear the special abort context for next time.
*/
if (TransactionAbortContext != NULL)
MemoryContextResetAndDeleteChildren(TransactionAbortContext);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
* Release all transaction-local memory.
*/
New memmgr logic in xact.c failed if AbortTransaction() is called when there is no open transaction.
2000-07-02 04:28:38 +02:00
if (TopTransactionContext != NULL)
MemoryContextDelete(TopTransactionContext);
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
TopTransactionContext = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CurTransactionContext = NULL;
CurrentTransactionState->curTransactionContext = NULL;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* ----------------------------------------------------------------
* CleanupSubTransaction stuff
* ----------------------------------------------------------------
*/
/*
* AtSubCleanup_Memory
*/
static void
AtSubCleanup_Memory(void)
{
TransactionState s = CurrentTransactionState;
Assert(s->parent != NULL);
/* Make sure we're not in an about-to-be-deleted context */
MemoryContextSwitchTo(s->parent->curTransactionContext);
CurTransactionContext = s->parent->curTransactionContext;
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/*
* Clear the special abort context for next time.
*/
if (TransactionAbortContext != NULL)
MemoryContextResetAndDeleteChildren(TransactionAbortContext);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* Delete the subxact local memory contexts. Its CurTransactionContext can
* go too (note this also kills CurTransactionContexts from any children
* of the subxact).
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (s->curTransactionContext)
MemoryContextDelete(s->curTransactionContext);
s->curTransactionContext = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
* interface routines
* ----------------------------------------------------------------
*/
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* StartTransaction
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
StartTransaction(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
TransactionState s;
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
VirtualTransactionId vxid;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* Let's just make sure the state stack is empty
*/
s = &TopTransactionStateData;
CurrentTransactionState = s;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
Assert(!FullTransactionIdIsValid(XactTopFullTransactionId));
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
Initialize TransactionState and user ID consistently at transaction start If a failure happens when a transaction is starting between the moment the transaction status is changed from TRANS_DEFAULT to TRANS_START and the moment the current user ID and security context flags are fetched via GetUserIdAndSecContext(), or before initializing its basic fields, then those may get reset to incorrect values when the transaction aborts, leaving the session in an inconsistent state. One problem reported is that failing a starting transaction at the first query of a session could cause several kinds of system crashes on the follow-up queries. In order to solve that, move the initialization of the transaction state fields and the call of GetUserIdAndSecContext() in charge of fetching the current user ID close to the point where the transaction status is switched to TRANS_START, where there cannot be any error triggered in-between, per an idea of Tom Lane. This properly ensures that the current user ID, the security context flags and that the basic fields of TransactionState remain consistent even if the transaction fails while starting. Reported-by: Richard Guo Diagnosed-By: Richard Guo Author: Michael Paquier Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAN_9JTxECSb=pEPcb0a8d+6J+bDcOZ4=DgRo_B7Y5gRHJUM=Rw@mail.gmail.com Backpatch-through: 9.4
2018-11-14 08:46:53 +01:00
/* check the current transaction state */
Assert(s->state == TRANS_DEFAULT);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Initialize TransactionState and user ID consistently at transaction start If a failure happens when a transaction is starting between the moment the transaction status is changed from TRANS_DEFAULT to TRANS_START and the moment the current user ID and security context flags are fetched via GetUserIdAndSecContext(), or before initializing its basic fields, then those may get reset to incorrect values when the transaction aborts, leaving the session in an inconsistent state. One problem reported is that failing a starting transaction at the first query of a session could cause several kinds of system crashes on the follow-up queries. In order to solve that, move the initialization of the transaction state fields and the call of GetUserIdAndSecContext() in charge of fetching the current user ID close to the point where the transaction status is switched to TRANS_START, where there cannot be any error triggered in-between, per an idea of Tom Lane. This properly ensures that the current user ID, the security context flags and that the basic fields of TransactionState remain consistent even if the transaction fails while starting. Reported-by: Richard Guo Diagnosed-By: Richard Guo Author: Michael Paquier Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAN_9JTxECSb=pEPcb0a8d+6J+bDcOZ4=DgRo_B7Y5gRHJUM=Rw@mail.gmail.com Backpatch-through: 9.4
2018-11-14 08:46:53 +01:00
* Set the current transaction state information appropriately during
* start processing. Note that once the transaction status is switched
* this process cannot fail until the user ID and the security context
* flags are fetched below.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
s->state = TRANS_START;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = InvalidFullTransactionId; /* until assigned */
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Log all statements from a sample of transactions This is useful to obtain a view of the different transaction types in an application, regardless of the durations of the statements each runs. Author: Adrien Nayrat Reviewed-by: Masahiko Sawada, Hayato Kuroda, Andres Freund
2019-04-03 23:43:59 +02:00
/* Determine if statements are logged in this transaction */
xact_is_sampled = log_xact_sample_rate != 0 &&
(log_xact_sample_rate == 1 ||
Replace random(), pg_erand48(), etc with a better PRNG API and algorithm. Standardize on xoroshiro128** as our basic PRNG algorithm, eliminating a bunch of platform dependencies as well as fundamentally-obsolete PRNG code. In addition, this API replacement will ease replacing the algorithm again in future, should that become necessary. xoroshiro128** is a few percent slower than the drand48 family, but it can produce full-width 64-bit random values not only 48-bit, and it should be much more trustworthy. It's likely to be noticeably faster than the platform's random(), depending on which platform you are thinking about; and we can have non-global state vectors easily, unlike with random(). It is not cryptographically strong, but neither are the functions it replaces. Fabien Coelho, reviewed by Dean Rasheed, Aleksander Alekseev, and myself Discussion: https://postgr.es/m/alpine.DEB.2.22.394.2105241211230.165418@pseudo
2021-11-29 03:32:36 +01:00
pg_prng_double(&pg_global_prng_state) <= log_xact_sample_rate);
Log all statements from a sample of transactions This is useful to obtain a view of the different transaction types in an application, regardless of the durations of the statements each runs. Author: Adrien Nayrat Reviewed-by: Masahiko Sawada, Hayato Kuroda, Andres Freund
2019-04-03 23:43:59 +02:00
Initialize TransactionState and user ID consistently at transaction start If a failure happens when a transaction is starting between the moment the transaction status is changed from TRANS_DEFAULT to TRANS_START and the moment the current user ID and security context flags are fetched via GetUserIdAndSecContext(), or before initializing its basic fields, then those may get reset to incorrect values when the transaction aborts, leaving the session in an inconsistent state. One problem reported is that failing a starting transaction at the first query of a session could cause several kinds of system crashes on the follow-up queries. In order to solve that, move the initialization of the transaction state fields and the call of GetUserIdAndSecContext() in charge of fetching the current user ID close to the point where the transaction status is switched to TRANS_START, where there cannot be any error triggered in-between, per an idea of Tom Lane. This properly ensures that the current user ID, the security context flags and that the basic fields of TransactionState remain consistent even if the transaction fails while starting. Reported-by: Richard Guo Diagnosed-By: Richard Guo Author: Michael Paquier Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAN_9JTxECSb=pEPcb0a8d+6J+bDcOZ4=DgRo_B7Y5gRHJUM=Rw@mail.gmail.com Backpatch-through: 9.4
2018-11-14 08:46:53 +01:00
/*
* initialize current transaction state fields
*
* note: prevXactReadOnly is not used at the outermost level
*/
s->nestingLevel = 1;
s->gucNestLevel = 1;
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
/*
* Once the current user ID and the security context flags are fetched,
* both will be properly reset even if transaction startup fails.
*/
GetUserIdAndSecContext(&s->prevUser, &s->prevSecContext);
/* SecurityRestrictionContext should never be set outside a transaction */
Assert(s->prevSecContext == 0);
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
/*
Improve snapshot manager by keeping explicit track of snapshots. There are two ways to track a snapshot: there's the "registered" list, which is used for arbitrary long-lived snapshots; and there's the "active stack", which is used for the snapshot that is considered "active" at any time. This also allows users of snapshots to stop worrying about snapshot memory allocation and freeing, and about using PG_TRY blocks around ActiveSnapshot assignment. This is all done automatically now. As a consequence, this allows us to reset MyProc->xmin when there are no more snapshots registered in the current backend, reducing the impact that long-running transactions have on VACUUM.
2008-05-12 22:02:02 +02:00
* Make sure we've reset xact state variables
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*
* If recovery is still in progress, mark this transaction as read-only.
* We have lower level defences in XLogInsert and elsewhere to stop us
* from modifying data during recovery, but this gives the normal
* indication to the user that the transaction is read-only.
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
*/
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
if (RecoveryInProgress())
{
s->startedInRecovery = true;
XactReadOnly = true;
}
else
{
s->startedInRecovery = false;
XactReadOnly = DefaultXactReadOnly;
}
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
XactDeferrable = DefaultXactDeferrable;
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
XactIsoLevel = DefaultXactIsoLevel;
Support an optional asynchronous commit mode, in which we don't flush WAL before reporting a transaction committed. Data consistency is still guaranteed (unlike setting fsync = off), but a crash may lose the effects of the last few transactions. Patch by Simon, some editorialization by Tom.
2007-08-02 00:45:09 +02:00
forceSyncCommit = false;
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
MyXactFlags = 0;
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* reinitialize within-transaction counters
*/
s->subTransactionId = TopSubTransactionId;
currentSubTransactionId = TopSubTransactionId;
currentCommandId = FirstCommandId;
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
currentCommandIdUsed = false;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* initialize reported xid accounting
*/
nUnreportedXids = 0;
Add new wal_level, logical, sufficient for logical decoding. When wal_level=logical, we'll log columns from the old tuple as configured by the REPLICA IDENTITY facility added in commit 07cacba983ef79be4a84fcd0e0ca3b5fcb85dd65. This makes it possible a properly-configured logical replication solution to correctly follow table updates even if they change the chosen key columns, or, with REPLICA IDENTITY FULL, even if the table has no key at all. Note that updates which do not modify the replica identity column won't log anything extra, making the choice of a good key (i.e. one that will rarely be changed) important to performance when wal_level=logical is configured. Each insert, update, or delete to a catalog table will also log the CMIN and/or CMAX values of stamped by the current transaction. This is necessary because logical decoding will require access to historical snapshots of the catalog in order to decode some data types, and the CMIN/CMAX values that we may need in order to judge row visibility may have been overwritten by the time we need them. Andres Freund, reviewed in various versions by myself, Heikki Linnakangas, KONDO Mitsumasa, and many others.
2013-12-11 00:33:45 +01:00
s->didLogXid = false;
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
* must initialize resource-management stuff first
*/
AtStart_Memory();
AtStart_ResourceOwner();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
* Assign a new LocalTransactionId, and combine it with the backendId to
* form a virtual transaction id.
*/
vxid.backendId = MyBackendId;
vxid.localTransactionId = GetNextLocalTransactionId();
/*
* Lock the virtual transaction id before we announce it in the proc array
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
VirtualXactLockTableInsert(vxid);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/*
* Advertise it in the proc array. We assume assignment of
Fix inconsistencies and typos in the tree This is numbered take 8, and addresses again a set of issues with code comments, variable names and unreferenced variables. Author: Alexander Lakhin Discussion: https://postgr.es/m/b137b5eb-9c95-9c2f-586e-38aba7d59788@gmail.com
2019-07-29 05:28:30 +02:00
* localTransactionId is atomic, and the backendId should be set already.
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
*/
Assert(MyProc->backendId == vxid.backendId);
MyProc->lxid = vxid.localTransactionId;
Initial MVCC code. New code for locking buffer' context.
1998-12-15 13:47:01 +01:00
Enable probes to work with Mac OS X Leopard and other OSes that will support DTrace in the future. Switch from using DTRACE_PROBEn macros to the dynamically generated macros. Use "dtrace -h" to create a header file that contains the dynamically generated macros to be used in the source code instead of the DTRACE_PROBEn macros. A dummy header file is generated for builds without DTrace support. Author: Robert Lor <Robert.Lor@sun.com>
2008-03-17 20:44:41 +01:00
TRACE_POSTGRESQL_TRANSACTION_START(vxid.localTransactionId);
DTrace support, with a small initial set of probes by Robert Lor
2006-07-24 18:32:45 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Advance transaction timestamp for intra-procedure transactions. Per discussion, this behavior seems less astonishing than not doing so. Peter Eisentraut and Tom Lane Discussion: https://postgr.es/m/20180920234040.GC29981@momjian.us
2018-10-08 22:16:36 +02:00
* set transaction_timestamp() (a/k/a now()). Normally, we want this to
* be the same as the first command's statement_timestamp(), so don't do a
* fresh GetCurrentTimestamp() call (which'd be expensive anyway). But
* for transactions started inside procedures (i.e., nonatomic SPI
* contexts), we do need to advance the timestamp. Also, in a parallel
* worker, the timestamp should already have been provided by a call to
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
* SetParallelStartTimestamps().
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
if (!IsParallelWorker())
Advance transaction timestamp for intra-procedure transactions. Per discussion, this behavior seems less astonishing than not doing so. Peter Eisentraut and Tom Lane Discussion: https://postgr.es/m/20180920234040.GC29981@momjian.us
2018-10-08 22:16:36 +02:00
{
if (!SPI_inside_nonatomic_context())
xactStartTimestamp = stmtStartTimestamp;
else
xactStartTimestamp = GetCurrentTimestamp();
}
Propagate xactStartTimestamp and stmtStartTimestamp to parallel workers. Previously, a worker process would establish values for these based on its own start time. In v10 and up, this can trivially be shown to cause misbehavior of transaction_timestamp(), timestamp_in(), and related functions which are (perhaps unwisely?) marked parallel-safe. It seems likely that other behaviors might diverge from what happens in the parent as well. It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure it's still possible, so back-patch to all branches containing parallel worker infrastructure. In HEAD only, mark now() and statement_timestamp() as parallel-safe (other affected functions already were). While in theory we could still squeeze that change into v11, it doesn't seem important enough to force a last-minute catversion bump. Konstantin Knizhnik, whacked around a bit by me Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
2018-10-06 18:00:09 +02:00
else
Assert(xactStartTimestamp != 0);
Rename recently-added pg_stat_activity column from txn_start to xact_start, for consistency with other column names such as in pg_stat_database.
2007-09-11 05:28:05 +02:00
pgstat_report_xact_timestamp(xactStartTimestamp);
Advance transaction timestamp for intra-procedure transactions. Per discussion, this behavior seems less astonishing than not doing so. Peter Eisentraut and Tom Lane Discussion: https://postgr.es/m/20180920234040.GC29981@momjian.us
2018-10-08 22:16:36 +02:00
/* Mark xactStopTimestamp as unset. */
xactStopTimestamp = 0;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
* initialize other subsystems for new transaction
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
AtStart_GUC();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
AtStart_Cache();
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
AfterTriggerBeginXact();
This is part #1 for of the DEFERRED CONSTRAINT TRIGGER support. Implements the CREATE CONSTRAINT TRIGGER and SET CONSTRAINTS commands. TODO: Generic builtin trigger procedures Automatic execution of appropriate CREATE CONSTRAINT... at CREATE TABLE Support of new trigger type in pg_dump Swapping of huge # of events to disk Jan
1999-09-29 18:06:40 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* done with start processing, set current transaction state to "in
* progress"
*/
Setting MyProc->xid to current transaction' id in StartTransaction and to InvalidTransactionId in CommitTransaction & AbortTransaction (it's for new TransactionIdIsInProgress func).
1996-11-27 08:14:51 +01:00
s->state = TRANS_INPROGRESS;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
ShowTransactionState("StartTransaction");
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* CommitTransaction
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*
* NB: if you change this routine, better look at PrepareTransaction too!
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Make functions static where possible, enclose unused functions in #ifdef NOT_USED.
1997-08-19 23:40:56 +02:00
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
CommitTransaction(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
TransactionId latestXid;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
bool is_parallel_worker;
is_parallel_worker = (s->blockState == TBLOCK_PARALLEL_INPROGRESS);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Tighten up application of parallel mode checks. Commit 924bcf4f16d54c55310b28f77686608684734f42 failed to enforce parallel mode checks during the commit of a parallel worker, because we exited parallel mode prior to ending the transaction so that we could pop the active snapshot. Re-establish parallel mode during parallel worker commit. Without this, it's far too easy for unsafe actions during the pre-commit sequence to crash the server instead of hitting the error checks as intended. Just to be extra paranoid, adjust a couple of the sanity checks in xact.c to check not only IsInParallelMode() but also IsParallelWorker().
2015-10-16 15:59:57 +02:00
/* Enforce parallel mode restrictions during parallel worker commit. */
if (is_parallel_worker)
EnterParallelMode();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
ShowTransactionState("CommitTransaction");
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* check the current transaction state
*/
if (s->state != TRANS_INPROGRESS)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "CommitTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Assert(s->parent == NULL);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
/*
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
* Do pre-commit processing that involves calling user-defined code, such
In security-restricted operations, block enqueue of at-commit user code. Specifically, this blocks DECLARE ... WITH HOLD and firing of deferred triggers within index expressions and materialized view queries. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. One can work around the vulnerability by disabling autovacuum and not manually running ANALYZE, CLUSTER, REINDEX, CREATE INDEX, VACUUM FULL, or REFRESH MATERIALIZED VIEW. (Don't restore from pg_dump, since it runs some of those commands.) Plain VACUUM (without FULL) is safe, and all commands are fine when a trusted user owns the target object. Performance may degrade quickly under this workaround, however. Back-patch to 9.5 (all supported versions). Reviewed by Robert Haas. Reported by Etienne Stalmans. Security: CVE-2020-25695
2020-11-09 16:32:09 +01:00
* as triggers. SECURITY_RESTRICTED_OPERATION contexts must not queue an
* action that would run here, because that would bypass the sandbox.
* Since closing cursors could queue trigger actions, triggers could open
* cursors, etc, we have to keep looping until there's nothing left to do.
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
*/
Fix interaction between materializing holdable cursors and firing deferred triggers: either one can create more work for the other, so we have to loop till it's all gone. Per example from andrew@supernews. Add a regression test to help spot trouble in this area in future.
2005-04-11 21:51:16 +02:00
for (;;)
{
/*
* Fire all currently pending deferred triggers.
*/
AfterTriggerFireDeferred();
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
Fix interaction between materializing holdable cursors and firing deferred triggers: either one can create more work for the other, so we have to loop till it's all gone. Per example from andrew@supernews. Add a regression test to help spot trouble in this area in future.
2005-04-11 21:51:16 +02:00
/*
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
* Close open portals (converting holdable ones into static portals).
* If there weren't any, we are done ... otherwise loop back to check
* if they queued deferred triggers. Lather, rinse, repeat.
Fix interaction between materializing holdable cursors and firing deferred triggers: either one can create more work for the other, so we have to loop till it's all gone. Per example from andrew@supernews. Add a regression test to help spot trouble in this area in future.
2005-04-11 21:51:16 +02:00
*/
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
if (!PreCommit_Portals(false))
Fix interaction between materializing holdable cursors and firing deferred triggers: either one can create more work for the other, so we have to loop till it's all gone. Per example from andrew@supernews. Add a regression test to help spot trouble in this area in future.
2005-04-11 21:51:16 +02:00
break;
}
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
/*
* The remaining actions cannot call any user-defined code, so it's safe
* to start shutting down within-transaction services. But note that most
* of this stuff could still throw an error, which would switch us into
* the transaction-abort path.
*/
This is part #1 for of the DEFERRED CONSTRAINT TRIGGER support. Implements the CREATE CONSTRAINT TRIGGER and SET CONSTRAINTS commands. TODO: Generic builtin trigger procedures Automatic execution of appropriate CREATE CONSTRAINT... at CREATE TABLE Support of new trigger type in pg_dump Swapping of huge # of events to disk Jan
1999-09-29 18:06:40 +02:00
In security-restricted operations, block enqueue of at-commit user code. Specifically, this blocks DECLARE ... WITH HOLD and firing of deferred triggers within index expressions and materialized view queries. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. One can work around the vulnerability by disabling autovacuum and not manually running ANALYZE, CLUSTER, REINDEX, CREATE INDEX, VACUUM FULL, or REFRESH MATERIALIZED VIEW. (Don't restore from pg_dump, since it runs some of those commands.) Plain VACUUM (without FULL) is safe, and all commands are fine when a trusted user owns the target object. Performance may degrade quickly under this workaround, however. Back-patch to 9.5 (all supported versions). Reviewed by Robert Haas. Reported by Etienne Stalmans. Security: CVE-2020-25695
2020-11-09 16:32:09 +01:00
CallXactCallbacks(is_parallel_worker ? XACT_EVENT_PARALLEL_PRE_COMMIT
: XACT_EVENT_PRE_COMMIT);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/* If we might have parallel workers, clean them up now. */
if (IsInParallelMode())
AtEOXact_Parallel(true);
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
/* Shut down the deferred-trigger manager */
AfterTriggerEndXact(true);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
* Let ON COMMIT management do its thing (must happen after closing
* cursors, to avoid dangling-reference problems)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
PreCommit_on_commit_actions();
Make CREATE/ALTER/DROP USER/GROUP transaction-safe, or at least pretty nearly so, by postponing write of flat password file until transaction commit.
2002-10-21 21:46:45 +02:00
Skip WAL for new relfilenodes, under wal_level=minimal. Until now, only selected bulk operations (e.g. COPY) did this. If a given relfilenode received both a WAL-skipping COPY and a WAL-logged operation (e.g. INSERT), recovery could lose tuples from the COPY. See src/backend/access/transam/README section "Skipping WAL for New RelFileNode" for the new coding rules. Maintainers of table access methods should examine that section. To maintain data durability, just before commit, we choose between an fsync of the relfilenode and copying its contents to WAL. A new GUC, wal_skip_threshold, guides that choice. If this change slows a workload that creates small, permanent relfilenodes under wal_level=minimal, try adjusting wal_skip_threshold. Users setting a timeout on COMMIT may need to adjust that timeout, and log_min_duration_statement analysis will reflect time consumption moving to COMMIT from commands like COPY. Internally, this requires a reliable determination of whether RollbackAndReleaseCurrentSubTransaction() would unlink a relation's current relfilenode. Introduce rd_firstRelfilenodeSubid. Amend the specification of rd_createSubid such that the field is zero when a new rel has an old rd_node. Make relcache.c retain entries for certain dropped relations until end of transaction. Bump XLOG_PAGE_MAGIC, since this introduces XLOG_GIST_ASSIGN_LSN. Future servers accept older WAL, so this bump is discretionary. Kyotaro Horiguchi, reviewed (in earlier, similar versions) by Robert Haas. Heikki Linnakangas and Michael Paquier implemented earlier designs that materially clarified the problem. Reviewed, in earlier designs, by Andrew Dunstan, Andres Freund, Alvaro Herrera, Tom Lane, Fujii Masao, and Simon Riggs. Reported by Martijn van Oosterhout. Discussion: https://postgr.es/m/20150702220524.GA9392@svana.org
2020-04-04 21:25:34 +02:00
/*
* Synchronize files that are created and not WAL-logged during this
* transaction. This must happen before AtEOXact_RelationMap(), so that we
* don't see committed-but-broken files after a crash.
*/
smgrDoPendingSyncs(true, is_parallel_worker);
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
/* close large objects before lower-level cleanup */
AtEOXact_LargeObject(true);
Theses buffer leaks are caused by indexes that are kept open between calls. Outside a transaction, the backend detects them as buffer leaks; it sends a NOTICE, and frees them. This sometimes cause a segmentation fault (at least on Linux). These indexes are initialized on the first lo_read/lo_write/lo_tell call, and (normally) closed on a lo_close call. Thus the buffer leaks appear when lo direct access functions are used, and not with lo_import/lo_export functions (libpq version calls lo_close before ending the command, and the backend version uses another path). The included patches (against recent snapshot, and against 6.3.2) cause indexes to be closed on transaction end (that is on explicit 'END' statment, or on command termination outside trasaction blocks), thus preventing the buffer leaks while increasing performance inside transactions. Some (all?) 'classic' memory leaks are also removed. I hope it will be ok. --- Pascal ANDRE, graduated from Ecole Centrale Paris andre@via.ecp.fr
1998-07-21 06:17:30 +02:00
Avoid assertion failure with LISTEN in a serializable transaction. If LISTEN is the only action in a serializable-mode transaction, and the session was not previously listening, and the notify queue is not empty, predicate.c reported an assertion failure. That happened because we'd acquire the transaction's initial snapshot during PreCommit_Notify, which was called *after* predicate.c expects any such snapshot to have been established. To fix, just swap the order of the PreCommit_Notify and PreCommit_CheckForSerializationFailure calls during CommitTransaction. This will imply holding the notify-insertion lock slightly longer, but the difference could only be meaningful in serializable mode, which is an expensive option anyway. It appears that this is just an assertion failure, with no consequences in non-assert builds. A snapshot used only to scan the notify queue could not have been involved in any serialization conflicts, so there would be nothing for PreCommit_CheckForSerializationFailure to do except assign it a prepareSeqNo and set the SXACT_FLAG_PREPARED flag. And given no conflicts, neither of those omissions affect the behavior of ReleasePredicateLocks. This admittedly once-over-lightly analysis is backed up by the lack of field reports of trouble. Per report from Mark Dilger. The bug is old, so back-patch to all supported branches; but the new test case only goes back to 9.6, for lack of adequate isolationtester infrastructure before that. Discussion: https://postgr.es/m/3ac7f397-4d5f-be8e-f354-440020675694@gmail.com Discussion: https://postgr.es/m/13881.1574557302@sss.pgh.pa.us
2019-11-24 21:57:31 +01:00
/*
* Insert notifications sent by NOTIFY commands into the queue. This
* should be late in the pre-commit sequence to minimize time spent
* holding the notify-insertion lock. However, this could result in
* creating a snapshot, so we must do it before serializable cleanup.
*/
PreCommit_Notify();
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
/*
* Mark serializable transaction as complete for predicate locking
* purposes. This should be done as late as we can put it and still allow
Enable parallel query with SERIALIZABLE isolation. Previously, the SERIALIZABLE isolation level prevented parallel query from being used. Allow the two features to be used together by sharing the leader's SERIALIZABLEXACT with parallel workers. An extra per-SERIALIZABLEXACT LWLock is introduced to make it safe to share, and new logic is introduced to coordinate the early release of the SERIALIZABLEXACT required for the SXACT_FLAG_RO_SAFE optimization, as follows: The first backend to observe the SXACT_FLAG_RO_SAFE flag (set by some other transaction) will 'partially release' the SERIALIZABLEXACT, meaning that the conflicts and locks it holds are released, but the SERIALIZABLEXACT itself will remain active because other backends might still have a pointer to it. Whenever any backend notices the SXACT_FLAG_RO_SAFE flag, it clears its own MySerializableXact variable and frees local resources so that it can skip SSI checks for the rest of the transaction. In the special case of the leader process, it transfers the SERIALIZABLEXACT to a new variable SavedSerializableXact, so that it can be completely released at the end of the transaction after all workers have exited. Remove the serializable_okay flag added to CreateParallelContext() by commit 9da0cc35, because it's now redundant. Author: Thomas Munro Reviewed-by: Haribabu Kommi, Robert Haas, Masahiko Sawada, Kevin Grittner Discussion: https://postgr.es/m/CAEepm=0gXGYhtrVDWOTHS8SQQy_=S9xo+8oCxGLWZAOoeJ=yzQ@mail.gmail.com
2019-03-15 04:23:46 +01:00
* errors to be raised for failure patterns found at commit. This is not
* appropriate in a parallel worker however, because we aren't committing
* the leader's transaction and its serializable state will live on.
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
*/
Enable parallel query with SERIALIZABLE isolation. Previously, the SERIALIZABLE isolation level prevented parallel query from being used. Allow the two features to be used together by sharing the leader's SERIALIZABLEXACT with parallel workers. An extra per-SERIALIZABLEXACT LWLock is introduced to make it safe to share, and new logic is introduced to coordinate the early release of the SERIALIZABLEXACT required for the SXACT_FLAG_RO_SAFE optimization, as follows: The first backend to observe the SXACT_FLAG_RO_SAFE flag (set by some other transaction) will 'partially release' the SERIALIZABLEXACT, meaning that the conflicts and locks it holds are released, but the SERIALIZABLEXACT itself will remain active because other backends might still have a pointer to it. Whenever any backend notices the SXACT_FLAG_RO_SAFE flag, it clears its own MySerializableXact variable and frees local resources so that it can skip SSI checks for the rest of the transaction. In the special case of the leader process, it transfers the SERIALIZABLEXACT to a new variable SavedSerializableXact, so that it can be completely released at the end of the transaction after all workers have exited. Remove the serializable_okay flag added to CreateParallelContext() by commit 9da0cc35, because it's now redundant. Author: Thomas Munro Reviewed-by: Haribabu Kommi, Robert Haas, Masahiko Sawada, Kevin Grittner Discussion: https://postgr.es/m/CAEepm=0gXGYhtrVDWOTHS8SQQy_=S9xo+8oCxGLWZAOoeJ=yzQ@mail.gmail.com
2019-03-15 04:23:46 +01:00
if (!is_parallel_worker)
PreCommit_CheckForSerializationFailure();
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
/* Prevent cancel/die interrupt while cleaning up */
HOLD_INTERRUPTS();
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
/* Commit updates to the relation map --- do this as late as possible */
Handle parallel index builds on mapped relations. Commit 9da0cc35284, which introduced parallel CREATE INDEX, failed to propagate relmapper.c backend local cache state to parallel worker processes. This could result in parallel index builds against mapped catalog relations where the leader process (participating as a worker) scans the new, pristine relfilenode, while worker processes scan the obsolescent relfilenode. When this happened, the final index structure was typically not consistent with the owning table's structure. The final index structure could contain entries formed from both heap relfilenodes. Only rebuilds on mapped catalog relations that occur as part of a VACUUM FULL or CLUSTER could become corrupt in practice, since their mapped relation relfilenode swap is what allows the inconsistency to arise. On master, fix the problem by propagating the required relmapper.c backend state as part of standard parallel initialization (Cf. commit 29d58fd3). On v11, simply disallow builds against mapped catalog relations by deeming them parallel unsafe. Author: Peter Geoghegan Reported-By: "death lock" Reviewed-By: Tom Lane, Amit Kapila Bug: #15309 Discussion: https://postgr.es/m/153329671686.1405.18298309097348420351@wrigleys.postgresql.org Backpatch: 11-, where parallel CREATE INDEX was introduced.
2018-08-10 22:01:34 +02:00
AtEOXact_RelationMap(true, is_parallel_worker);
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
/*
* set the current transaction state information appropriately during
Fix a couple of bogus comments, per Alvaro.
2005-07-14 00:46:09 +02:00
* commit processing
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
*/
s->state = TRANS_COMMIT;
Tighten up application of parallel mode checks. Commit 924bcf4f16d54c55310b28f77686608684734f42 failed to enforce parallel mode checks during the commit of a parallel worker, because we exited parallel mode prior to ending the transaction so that we could pop the active snapshot. Re-establish parallel mode during parallel worker commit. Without this, it's far too easy for unsafe actions during the pre-commit sequence to crash the server instead of hitting the error checks as intended. Just to be extra paranoid, adjust a couple of the sanity checks in xact.c to check not only IsInParallelMode() but also IsParallelWorker().
2015-10-16 15:59:57 +02:00
s->parallelModeLevel = 0;
Rearrange order of pre-commit operations: must close cursors before doing ON COMMIT actions. Per bug report from Michael Guerin.
2004-10-30 00:19:53 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
if (!is_parallel_worker)
{
/*
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
* We need to mark our XIDs as committed in pg_xact. This is where we
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* durably commit.
*/
latestXid = RecordTransactionCommit();
}
else
{
/*
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* We must not mark our XID committed; the parallel leader is
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* responsible for that.
*/
latestXid = InvalidTransactionId;
/*
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* Make sure the leader will know about any WAL we wrote before it
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* commits.
*/
ParallelWorkerReportLastRecEnd(XactLastRecEnd);
}
1. Additional fix against ERROR: Child itemid marked as unused in CommitTransaction(). 2. Changes in GetSnapshotData().
1999-06-03 15:33:13 +02:00
Enable probes to work with Mac OS X Leopard and other OSes that will support DTrace in the future. Switch from using DTRACE_PROBEn macros to the dynamically generated macros. Use "dtrace -h" to create a header file that contains the dynamically generated macros to be used in the source code instead of the DTRACE_PROBEn macros. A dummy header file is generated for builds without DTrace support. Author: Robert Lor <Robert.Lor@sun.com>
2008-03-17 20:44:41 +01:00
TRACE_POSTGRESQL_TRANSACTION_COMMIT(MyProc->lxid);
Don't take ProcArrayLock while exiting a transaction that has no XID; there is no need for serialization against snapshot-taking because the xact doesn't affect anyone else's snapshot anyway. Per discussion. Also, move various info about the interlocking of transactions and snapshots out of code comments and into a hopefully-more-cohesive discussion in access/transam/README. Also, remove a couple of now-obsolete comments about having to force some WAL to be written to persuade RecordTransactionCommit to do its thing.
2007-09-07 22:59:26 +02:00
/*
1. Additional fix against ERROR: Child itemid marked as unused in CommitTransaction(). 2. Changes in GetSnapshotData().
1999-06-03 15:33:13 +02:00
* Let others know about no transaction in progress by me. Note that this
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
* must be done _before_ releasing locks we hold and _after_
* RecordTransactionCommit.
1. Additional fix against ERROR: Child itemid marked as unused in CommitTransaction(). 2. Changes in GetSnapshotData().
1999-06-03 15:33:13 +02:00
*/
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
ProcArrayEndTransaction(MyProc, latestXid);
1. Additional fix against ERROR: Child itemid marked as unused in CommitTransaction(). 2. Changes in GetSnapshotData().
1999-06-03 15:33:13 +02:00
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
/*
* This is all post-commit cleanup. Note that if an error is raised here,
* it's too late to abort the transaction. This should be just
* noncritical resource releasing.
Perform transaction cleanup operations in a less ad-hoc, more principled order; in particular ensure that all shared resources are released before we release transaction locks. The code used to release locks before buffer pins, which might explain an ancient note I have about a bufmgr assertion failure I'd seen once several years ago, and been unable to reproduce since. (Theory: someone trying to drop a relation might be able to reach FlushRelationBuffers before the last user of the relation had gotten around to dropping his buffer pins.)
2002-10-23 00:44:36 +02:00
*
* The ordering of operations is not entirely random. The idea is:
* release resources visible to other backends (eg, files, buffer pins);
* then release locks; then release backend-local resources. We want to
* release locks at the point where any backend waiting for us will see
* our transaction as being fully cleaned up.
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
*
* Resources that can be associated with individual queries are handled by
* the ResourceOwner mechanism. The other calls here are for backend-wide
* state.
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
*/
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
CallXactCallbacks(is_parallel_worker ? XACT_EVENT_PARALLEL_COMMIT
: XACT_EVENT_COMMIT);
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
true, true);
Give the ResourceOwner mechanism full responsibility for releasing buffer pins at end of transaction, and reduce AtEOXact_Buffers to an Assert cross-check that this was done correctly. When not USE_ASSERT_CHECKING, AtEOXact_Buffers is a complete no-op. This gets rid of an O(NBuffers) bottleneck during transaction commit/abort, which recent testing has shown becomes significant above a few tens of thousands of shared buffers.
2004-10-16 20:57:26 +02:00
/* Check we've released all buffer pins */
AtEOXact_Buffers(true);
Modify AtEOXact_CatCache and AtEOXact_RelationCache to assume that the ResourceOwner mechanism already released all reference counts for the cache entries; therefore, we do not need to scan the catcache or relcache at transaction end, unless we want to do it as a debugging crosscheck. Do the crosscheck only in Assert mode. This is the same logic we had previously installed in AtEOXact_Buffers to avoid overhead with large numbers of shared buffers. I thought it'd be a good idea to do it here too, in view of Kari Lavikka's recent report showing a real-world case where AtEOXact_CatCache is taking a significant fraction of runtime.
2005-08-08 21:17:23 +02:00
/* Clean up the relation cache */
AtEOXact_RelationCache(true);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
* Make catalog changes visible to all backends. This has to happen after
* relcache references are dropped (see comments for
* AtEOXact_RelationCache), but before locks are released (if anyone is
* waiting for lock on a relation we've modified, we want them to know
* about the catalog change before they start using the relation).
*/
AtEOXact_Inval(true);
Implement sharable row-level locks, and use them for foreign key references to eliminate unnecessary deadlocks. This commit adds SELECT ... FOR SHARE paralleling SELECT ... FOR UPDATE. The implementation uses a new SLRU data structure (managed much like pg_subtrans) to represent multiple- transaction-ID sets. When more than one transaction is holding a shared lock on a particular row, we create a MultiXactId representing that set of transactions and store its ID in the row's XMAX. This scheme allows an effectively unlimited number of row locks, just as we did before, while not costing any extra overhead except when a shared lock actually has to be shared. Still TODO: use the regular lock manager to control the grant order when multiple backends are waiting for a row lock. Alvaro Herrera and Tom Lane.
2005-04-28 23:47:18 +02:00
AtEOXact_MultiXact();
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_LOCKS,
true, true);
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_AFTER_LOCKS,
true, true);
Make DROP TABLE rollback-able: postpone physical file delete until commit. (WAL logging for this is not done yet, however.) Clean up a number of really crufty things that are no longer needed now that DROP behaves nicely. Make temp table mapper do the right things when drop or rename affecting a temp table is rolled back. Also, remove "relation modified while in use" error check, in favor of locking tables at first reference and holding that lock throughout the statement.
2000-11-08 23:10:03 +01:00
During transaction cleanup, release locks before deleting files. There's no need to hold onto the locks until the files are needed, and by doing it this way, we reduce the impact on other backends who may be awaiting locks we hold. Noah Misch
2012-06-14 16:19:33 +02:00
/*
* Likewise, dropping of files deleted during the transaction is best done
* after releasing relcache and buffer pins. (This is not strictly
* necessary during commit, since such pins should have been released
* already, but this ordering is definitely critical during abort.) Since
* this may take many seconds, also delay until after releasing locks.
* Other backends will observe the attendant catalog changes and not
* attempt to access affected files.
*/
smgrDoPendingDeletes(true);
Send NOTIFY signals during CommitTransaction. Formerly, we sent signals for outgoing NOTIFY messages within ProcessCompletedNotifies, which was also responsible for sending relevant ones of those messages to our connected client. It therefore had to run during the main-loop processing that occurs just before going idle. This arrangement had two big disadvantages: * Now that procedures allow intra-command COMMITs, it would be useful to send NOTIFYs to other sessions immediately at COMMIT (though, for reasons of wire-protocol stability, we still shouldn't forward them to our client until end of command). * Background processes such as replication workers would not send NOTIFYs at all, since they never execute the client communication loop. We've had requests to allow triggers running in replication workers to send NOTIFYs, so that's a problem. To fix these things, move transmission of outgoing NOTIFY signals into AtCommit_Notify, where it will happen during CommitTransaction. Also move the possible call of asyncQueueAdvanceTail there, to ensure we don't bloat the async SLRU if a background worker sends many NOTIFYs with no one listening. We can also drop the call of asyncQueueReadAllNotifications, allowing ProcessCompletedNotifies to go away entirely. That's because commit 790026972 added a call of ProcessNotifyInterrupt adjacent to PostgresMain's call of ProcessCompletedNotifies, and that does its own call of asyncQueueReadAllNotifications, meaning that we were uselessly doing two such calls (inside two separate transactions) whenever inbound notify signals coincided with an outbound notify. We need only set notifyInterruptPending to ensure that ProcessNotifyInterrupt runs, and we're done. The existing documentation suggests that custom background workers should call ProcessCompletedNotifies if they want to send NOTIFY messages. To avoid an ABI break in the back branches, reduce it to an empty routine rather than removing it entirely. Removal will occur in v15. Although the problems mentioned above have existed for awhile, I don't feel comfortable back-patching this any further than v13. There was quite a bit of churn in adjacent code between 12 and 13. At minimum we'd have to also backpatch 51004c717, and a good deal of other adjustment would also be needed, so the benefit-to-risk ratio doesn't look attractive. Per bug #15293 from Michael Powers (and similar gripes from others). Artur Zakirov and Tom Lane Discussion: https://postgr.es/m/153243441449.1404.2274116228506175596@wrigleys.postgresql.org
2021-09-14 23:18:25 +02:00
/*
* Send out notification signals to other backends (and do other
* post-commit NOTIFY cleanup). This must not happen until after our
* transaction is fully done from the viewpoint of other backends.
*/
Replace the pg_listener-based LISTEN/NOTIFY mechanism with an in-memory queue. In addition, add support for a "payload" string to be passed along with each notify event. This implementation should be significantly more efficient than the old one, and is also more compatible with Hot Standby usage. There is not yet any facility for HS slaves to receive notifications generated on the master, although such a thing is possible in future. Joachim Wieland, reviewed by Jeff Davis; also hacked on by me.
2010-02-16 23:34:57 +01:00
AtCommit_Notify();
Send NOTIFY signals during CommitTransaction. Formerly, we sent signals for outgoing NOTIFY messages within ProcessCompletedNotifies, which was also responsible for sending relevant ones of those messages to our connected client. It therefore had to run during the main-loop processing that occurs just before going idle. This arrangement had two big disadvantages: * Now that procedures allow intra-command COMMITs, it would be useful to send NOTIFYs to other sessions immediately at COMMIT (though, for reasons of wire-protocol stability, we still shouldn't forward them to our client until end of command). * Background processes such as replication workers would not send NOTIFYs at all, since they never execute the client communication loop. We've had requests to allow triggers running in replication workers to send NOTIFYs, so that's a problem. To fix these things, move transmission of outgoing NOTIFY signals into AtCommit_Notify, where it will happen during CommitTransaction. Also move the possible call of asyncQueueAdvanceTail there, to ensure we don't bloat the async SLRU if a background worker sends many NOTIFYs with no one listening. We can also drop the call of asyncQueueReadAllNotifications, allowing ProcessCompletedNotifies to go away entirely. That's because commit 790026972 added a call of ProcessNotifyInterrupt adjacent to PostgresMain's call of ProcessCompletedNotifies, and that does its own call of asyncQueueReadAllNotifications, meaning that we were uselessly doing two such calls (inside two separate transactions) whenever inbound notify signals coincided with an outbound notify. We need only set notifyInterruptPending to ensure that ProcessNotifyInterrupt runs, and we're done. The existing documentation suggests that custom background workers should call ProcessCompletedNotifies if they want to send NOTIFY messages. To avoid an ABI break in the back branches, reduce it to an empty routine rather than removing it entirely. Removal will occur in v15. Although the problems mentioned above have existed for awhile, I don't feel comfortable back-patching this any further than v13. There was quite a bit of churn in adjacent code between 12 and 13. At minimum we'd have to also backpatch 51004c717, and a good deal of other adjustment would also be needed, so the benefit-to-risk ratio doesn't look attractive. Per bug #15293 from Michael Powers (and similar gripes from others). Artur Zakirov and Tom Lane Discussion: https://postgr.es/m/153243441449.1404.2274116228506175596@wrigleys.postgresql.org
2021-09-14 23:18:25 +02:00
/*
* Everything after this should be purely internal-to-this-backend
* cleanup.
*/
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
AtEOXact_GUC(true, 1);
Add a warning to AtEOXact_SPI() to catch cases where the current transaction has been committed without SPI_finish() being called first. Per recent discussion here: http://archives.postgresql.org/pgsql-patches/2003-11/msg00286.php
2003-12-02 20:26:47 +01:00
AtEOXact_SPI(true);
Relax transactional restrictions on ALTER TYPE ... ADD VALUE (redux). Originally committed as 15bc038f (plus some follow-ups), this was reverted in 28e07270 due to a problem discovered in parallel workers. This new version corrects that problem by sending the list of uncommitted enum values to parallel workers. Here follows the original commit message describing the change: To prevent possibly breaking indexes on enum columns, we must keep uncommitted enum values from getting stored in tables, unless we can be sure that any such column is new in the current transaction. Formerly, we enforced this by disallowing ALTER TYPE ... ADD VALUE from being executed at all in a transaction block, unless the target enum type had been created in the current transaction. This patch removes that restriction, and instead insists that an uncommitted enum value can't be referenced unless it belongs to an enum type created in the same transaction as the value. Per discussion, this should be a bit less onerous. It does require each function that could possibly return a new enum value to SQL operations to check this restriction, but there aren't so many of those that this seems unmaintainable. Author: Andrew Dunstan and Tom Lane, with parallel query fix by Thomas Munro Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAEepm%3D0Ei7g6PaNTbcmAh9tCRahQrk%3Dr5ZWLD-jr7hXweYX3yg%40mail.gmail.com Discussion: https://postgr.es/m/4075.1459088427%40sss.pgh.pa.us
2018-10-09 01:51:01 +02:00
AtEOXact_Enum();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtEOXact_on_commit_actions(true);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
AtEOXact_Namespace(true, is_parallel_worker);
Close un-owned SMgrRelations at transaction end. If an SMgrRelation is not "owned" by a relcache entry, don't allow it to live past transaction end. This design allows the same SMgrRelation to be used for blind writes of multiple blocks during a transaction, but ensures that we don't hold onto such an SMgrRelation indefinitely. Because an SMgrRelation typically corresponds to open file descriptors at the fd.c level, leaving it open when there's no corresponding relcache entry can mean that we prevent the kernel from reclaiming deleted disk space. (While CacheInvalidateSmgr messages usually fix that, there are cases where they're not issued, such as DROP DATABASE. We might want to add some more sinval messaging for that, but I'd be inclined to keep this type of logic anyway, since allowing VFDs to accumulate indefinitely for blind-written relations doesn't seem like a good idea.) This code replaces a previous attempt towards the same goal that proved to be unreliable. Back-patch to 9.1 where the previous patch was added.
2012-10-17 18:38:21 +02:00
AtEOXact_SMgr();
In AtEOXact_Files, complain if any files remain unclosed at commit. This change makes this module act more like most of our other low-level resource management modules. It's a caller error if something is not explicitly closed by the end of a successful transaction, so issue a WARNING about it. This would not actually have caught the file leak bug fixed in commit 231bcd080, because that was in a transaction-abort path; but it still seems like a good, and pretty cheap, cross-check. Discussion: https://postgr.es/m/152056616579.4966.583293218357089052@wrigleys.postgresql.org
2018-04-28 23:45:02 +02:00
AtEOXact_Files(true);
Combine cmin and cmax fields of HeapTupleHeaders into a single field, by keeping private state in each backend that has inserted and deleted the same tuple during its current top-level transaction. This is sufficient since there is no need to be able to determine the cmin/cmax from any other transaction. This gets us back down to 23-byte headers, removing a penalty paid in 8.0 to support subtransactions. Patch by Heikki Linnakangas, with minor revisions by moi, following a design hashed out awhile back on the pghackers list.
2007-02-09 04:35:35 +01:00
AtEOXact_ComboCid();
Fix dynahash.c to suppress hash bucket splits while a hash_seq_search() scan is in progress on the same hashtable. This seems the least invasive way to fix the recently-recognized problem that a split could cause the scan to visit entries twice or (with much lower probability) miss them entirely. The only field-reported problem caused by this is the "failed to re-find shared lock object" PANIC in COMMIT PREPARED reported by Michel Dorochevsky, which was caused by multiply visited entries. However, it seems certain that mdsync() is vulnerable to missing required fsync's due to missed entries, and I am fearful that RelationCacheInitializePhase2() might be at risk as well. Because of that and the generalized hazard presented by this bug, back-patch all the supported branches. Along the way, fix pg_prepared_statement() and pg_cursor() to not assume that the hashtables they are examining will stay static between calls. This is risky regardless of the newly noted dynahash problem, because hash_seq_search() has never promised to cope with deletion of table entries other than the just-returned one. There may be no bug here because the only supported way to call these functions is via ExecMakeTableFunctionResult() which will cycle them to completion before doing anything very interesting, but it seems best to get rid of the assumption. This affects 8.2 and HEAD only, since those functions weren't there earlier.
2007-04-27 01:24:46 +02:00
AtEOXact_HashTables(true);
Avoid counting transaction stats for parallel worker cooperating transaction. The transaction that is initiated by the parallel worker to cooperate with the actual transaction started by the main backend to complete the query execution should not be counted as a separate transaction. The other internal transactions started and committed by the parallel worker are still counted as separate transactions as we that is what we do in other places like autovacuum. This will partially fix the bloat in transaction stats due to additional transactions performed by parallel workers. For a complete fix, we need to decide how we want to show all the transactions that are started internally for various operations and that is a matter of separate patch. Reported-by: Haribabu Kommi Author: Haribabu Kommi Reviewed-by: Amit Kapila, Jamison Kirk and Rahila Syed Backpatch-through: 9.6 Discussion: https://postgr.es/m/CAJrrPGc9=jKXuScvNyQ+VNhO0FZk7LLAShAJRyZjnedd2D61EQ@mail.gmail.com
2019-04-10 04:54:15 +02:00
AtEOXact_PgStat(true, is_parallel_worker);
Avoid SnapshotResetXmin() during AtEOXact_Snapshot() For normal commits and aborts we already reset PgXact->xmin, so we can simply avoid running SnapshotResetXmin() twice. During performance tests by Alexander Korotkov, diagnosis by Andres Freund showed PgXact array as a bottleneck. After manual analysis by me of the code paths that touch those memory locations, I was able to identify extraneous code in the main transaction commit path. Avoiding touching highly contented shmem improves concurrent performance slightly on all workloads, confirmed by tests run by Ashutosh Sharma and Alexander Korotkov. Simon Riggs Discussion: CANP8+jJdXE9b+b9F8CQT-LuxxO0PBCB-SZFfMVAdp+akqo4zfg@mail.gmail.com
2017-04-06 14:31:52 +02:00
AtEOXact_Snapshot(true, false);
Fix logical replication launcher wake up and reset After the logical replication launcher was told to wake up at commit (for example, by a CREATE SUBSCRIPTION command), the flag to wake up was not reset, so it would be woken up at every following commit as well. So fix that by resetting the flag. Also, we don't need to wake up anything if the transaction was rolled back. Just reset the flag in that case. Author: Masahiko Sawada <sawada.mshk@gmail.com> Reported-by: Fujii Masao <masao.fujii@gmail.com>
2017-05-01 16:18:09 +02:00
AtEOXact_ApplyLauncher(true);
Wake up a subscription's replication worker processes after DDL. Waken related worker processes immediately at commit of a transaction that has performed ALTER SUBSCRIPTION (including the RENAME and OWNER variants). This reduces the response time for such operations. In the real world that might not be worth much, but it shaves several seconds off the runtime for the subscription test suite. In the case of PREPARE, we just throw away this notification state; it doesn't seem worth the work to preserve it. The workers will still react after the eventual COMMIT PREPARED, but not as quickly. Nathan Bossart Discussion: https://postgr.es/m/20221122004119.GA132961@nathanxps13
2023-01-06 22:08:20 +01:00
AtEOXact_LogicalRepWorkers(true);
Rename recently-added pg_stat_activity column from txn_start to xact_start, for consistency with other column names such as in pg_stat_database.
2007-09-11 05:28:05 +02:00
pgstat_report_xact_timestamp(0);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
CurrentResourceOwner = NULL;
ResourceOwnerDelete(TopTransactionResourceOwner);
s->curTransactionOwner = NULL;
CurTransactionResourceOwner = NULL;
TopTransactionResourceOwner = NULL;
Perform transaction cleanup operations in a less ad-hoc, more principled order; in particular ensure that all shared resources are released before we release transaction locks. The code used to release locks before buffer pins, which might explain an ancient note I have about a bufmgr assertion failure I'd seen once several years ago, and been unable to reproduce since. (Theory: someone trying to drop a relation might be able to reach FlushRelationBuffers before the last user of the relation had gotten around to dropping his buffer pins.)
2002-10-23 00:44:36 +02:00
AtCommit_Memory();
Statistical system views (yet without the config stuff, but it's hard to keep such massive changes in sync with the tree so I need to get it in and work from there now). Jan
2001-06-22 21:16:24 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = InvalidFullTransactionId;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->subTransactionId = InvalidSubTransactionId;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->nestingLevel = 0;
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
s->gucNestLevel = 0;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
XactTopFullTransactionId = InvalidFullTransactionId;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nParallelCurrentXids = 0;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* done with commit processing, set current transaction state back to
* default
*/
s->state = TRANS_DEFAULT;
Restructure backend SIGINT/SIGTERM handling so that 'die' interrupts are treated more like 'cancel' interrupts: the signal handler sets a flag that is examined at well-defined spots, rather than trying to cope with an interrupt that might happen anywhere. See pghackers discussion of 1/12/01.
2001-01-14 06:08:17 +01:00
Make critical sections (elog->crash) and interrupt holdoff sections into distinct concepts, per recent discussion on pghackers.
2001-01-19 23:08:47 +01:00
RESUME_INTERRUPTS();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* PrepareTransaction
*
* NB: if you change this routine, better look at CommitTransaction too!
*/
static void
PrepareTransaction(void)
{
TransactionState s = CurrentTransactionState;
TransactionId xid = GetCurrentTransactionId();
GlobalTransaction gxact;
Add a time-of-preparation column to the pg_prepared_xacts view, per an old suggestion by Oliver Jowett. Also, add a transaction column to the pg_locks view to show the xid of each transaction holding or awaiting locks; this allows prepared transactions to be properly associated with the locks they own. There was already a column named 'transaction', and I chose to rename it to 'transactionid' --- since this column is new in the current devel cycle there should be no backwards compatibility issue to worry about.
2005-06-18 21:33:42 +02:00
TimestampTz prepared_at;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
Assert(!IsInParallelMode());
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
ShowTransactionState("PrepareTransaction");
/*
* check the current transaction state
*/
if (s->state != TRANS_INPROGRESS)
elog(WARNING, "PrepareTransaction while in %s state",
TransStateAsString(s->state));
Assert(s->parent == NULL);
/*
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
* Do pre-commit processing that involves calling user-defined code, such
* as triggers. Since closing cursors could queue trigger actions,
* triggers could open cursors, etc, we have to keep looping until there's
* nothing left to do.
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*/
for (;;)
{
/*
* Fire all currently pending deferred triggers.
*/
AfterTriggerFireDeferred();
/*
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
* Close open portals (converting holdable ones into static portals).
* If there weren't any, we are done ... otherwise loop back to check
* if they queued deferred triggers. Lather, rinse, repeat.
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*/
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
if (!PreCommit_Portals(true))
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
break;
}
Invent pre-commit/pre-prepare/pre-subcommit events for xact callbacks. Currently it's only possible for loadable modules to get control during post-commit cleanup of a transaction. That doesn't work too well if they want to do something that could throw an error; for example, an FDW might need to issue a remote commit, which could well fail. To improve matters, extend the existing APIs for XactCallback and SubXactCallback functions to provide new pre-commit events for this purpose. The release notes will need to mention that existing callback functions should be checked to make sure they don't do something unwanted when one of the new event types occurs. In the examples within our source tree, contrib/sepgsql was fine but plpgsql had been a bit too cute.
2013-02-15 02:35:08 +01:00
CallXactCallbacks(XACT_EVENT_PRE_PREPARE);
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
/*
* The remaining actions cannot call any user-defined code, so it's safe
* to start shutting down within-transaction services. But note that most
* of this stuff could still throw an error, which would switch us into
* the transaction-abort path.
*/
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Refactor the executor's API to support data-modifying CTEs better. The originally committed patch for modifying CTEs didn't interact well with EXPLAIN, as noted by myself, and also had corner-case problems with triggers, as noted by Dean Rasheed. Those problems show it is really not practical for ExecutorEnd to call any user-defined code; so split the cleanup duties out into a new function ExecutorFinish, which must be called between the last ExecutorRun call and ExecutorEnd. Some Asserts have been added to these functions to help verify correct usage. It is no longer necessary for callers of the executor to call AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now done by ExecutorStart/ExecutorFinish respectively. If you really need to suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to ExecutorStart. Also, refactor portal commit processing to allow for the possibility that PortalDrop will invoke user-defined code. I think this is not actually necessary just yet, since the portal-execution-strategy logic forces any non-pure-SELECT query to be run to completion before we will consider committing. But it seems like good future-proofing.
2011-02-27 19:43:29 +01:00
/* Shut down the deferred-trigger manager */
AfterTriggerEndXact(true);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* Let ON COMMIT management do its thing (must happen after closing
* cursors, to avoid dangling-reference problems)
*/
PreCommit_on_commit_actions();
Skip WAL for new relfilenodes, under wal_level=minimal. Until now, only selected bulk operations (e.g. COPY) did this. If a given relfilenode received both a WAL-skipping COPY and a WAL-logged operation (e.g. INSERT), recovery could lose tuples from the COPY. See src/backend/access/transam/README section "Skipping WAL for New RelFileNode" for the new coding rules. Maintainers of table access methods should examine that section. To maintain data durability, just before commit, we choose between an fsync of the relfilenode and copying its contents to WAL. A new GUC, wal_skip_threshold, guides that choice. If this change slows a workload that creates small, permanent relfilenodes under wal_level=minimal, try adjusting wal_skip_threshold. Users setting a timeout on COMMIT may need to adjust that timeout, and log_min_duration_statement analysis will reflect time consumption moving to COMMIT from commands like COPY. Internally, this requires a reliable determination of whether RollbackAndReleaseCurrentSubTransaction() would unlink a relation's current relfilenode. Introduce rd_firstRelfilenodeSubid. Amend the specification of rd_createSubid such that the field is zero when a new rel has an old rd_node. Make relcache.c retain entries for certain dropped relations until end of transaction. Bump XLOG_PAGE_MAGIC, since this introduces XLOG_GIST_ASSIGN_LSN. Future servers accept older WAL, so this bump is discretionary. Kyotaro Horiguchi, reviewed (in earlier, similar versions) by Robert Haas. Heikki Linnakangas and Michael Paquier implemented earlier designs that materially clarified the problem. Reviewed, in earlier designs, by Andrew Dunstan, Andres Freund, Alvaro Herrera, Tom Lane, Fujii Masao, and Simon Riggs. Reported by Martijn van Oosterhout. Discussion: https://postgr.es/m/20150702220524.GA9392@svana.org
2020-04-04 21:25:34 +02:00
/*
* Synchronize files that are created and not WAL-logged during this
* transaction. This must happen before EndPrepare(), so that we don't see
* committed-but-broken files after a crash and COMMIT PREPARED.
*/
smgrDoPendingSyncs(true, false);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* close large objects before lower-level cleanup */
AtEOXact_LargeObject(true);
Avoid assertion failure with LISTEN in a serializable transaction. If LISTEN is the only action in a serializable-mode transaction, and the session was not previously listening, and the notify queue is not empty, predicate.c reported an assertion failure. That happened because we'd acquire the transaction's initial snapshot during PreCommit_Notify, which was called *after* predicate.c expects any such snapshot to have been established. To fix, just swap the order of the PreCommit_Notify and PreCommit_CheckForSerializationFailure calls during CommitTransaction. This will imply holding the notify-insertion lock slightly longer, but the difference could only be meaningful in serializable mode, which is an expensive option anyway. It appears that this is just an assertion failure, with no consequences in non-assert builds. A snapshot used only to scan the notify queue could not have been involved in any serialization conflicts, so there would be nothing for PreCommit_CheckForSerializationFailure to do except assign it a prepareSeqNo and set the SXACT_FLAG_PREPARED flag. And given no conflicts, neither of those omissions affect the behavior of ReleasePredicateLocks. This admittedly once-over-lightly analysis is backed up by the lack of field reports of trouble. Per report from Mark Dilger. The bug is old, so back-patch to all supported branches; but the new test case only goes back to 9.6, for lack of adequate isolationtester infrastructure before that. Discussion: https://postgr.es/m/3ac7f397-4d5f-be8e-f354-440020675694@gmail.com Discussion: https://postgr.es/m/13881.1574557302@sss.pgh.pa.us
2019-11-24 21:57:31 +01:00
/* NOTIFY requires no work at this point */
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
/*
* Mark serializable transaction as complete for predicate locking
* purposes. This should be done as late as we can put it and still allow
* errors to be raised for failure patterns found at commit.
*/
PreCommit_CheckForSerializationFailure();
Fix PREPARE TRANSACTION to reject the case where the transaction has dropped a temporary table; we can't support that because there's no way to clean up the source backend's internal state if the eventual COMMIT PREPARED is done by another backend. This was checked correctly in 8.1 but I broke it in 8.2 :-(. Patch by Heikki Linnakangas, original trouble report by John Smith.
2008-03-04 20:54:06 +01:00
/*
* Don't allow PREPARE TRANSACTION if we've accessed a temporary table in
* this transaction. Having the prepared xact hold locks on another
* backend's temp table seems a bad idea --- for instance it would prevent
* the backend from exiting. There are other problems too, such as how to
* clean up the source backend's local buffers and ON COMMIT state if the
* prepared xact includes a DROP of a temp table.
*
Simplify restriction handling of two-phase commit for temporary objects There were two flags used to track the access to temporary tables and to the temporary namespace of a session which are used to restrict PREPARE TRANSACTION, however the first control flag is a concept included in the second. This removes the flag for temporary table tracking, keeping around only the one at namespace level. Author: Michael Paquier Reviewed-by: Álvaro Herrera Discussion: https://postgr.es/m/20190118053126.GH1883@paquier.xyz
2019-01-26 02:45:23 +01:00
* Other objects types, like functions, operators or extensions, share the
* same restriction as they should not be created, locked or dropped as
* this can mess up with this session or even a follow-up session trying
* to use the same temporary namespace.
*
Fix PREPARE TRANSACTION to reject the case where the transaction has dropped a temporary table; we can't support that because there's no way to clean up the source backend's internal state if the eventual COMMIT PREPARED is done by another backend. This was checked correctly in 8.1 but I broke it in 8.2 :-(. Patch by Heikki Linnakangas, original trouble report by John Smith.
2008-03-04 20:54:06 +01:00
* We must check this after executing any ON COMMIT actions, because they
* might still access a temp relation.
*
* XXX In principle this could be relaxed to allow some useful special
* cases, such as a temp table created and dropped all within the
* transaction. That seems to require much more bookkeeping though.
*/
Restrict the use of temporary namespace in two-phase transactions Attempting to use a temporary table within a two-phase transaction is forbidden for ages. However, there have been uncovered grounds for a couple of other object types and commands which work on temporary objects with two-phase commit. In short, trying to create, lock or drop an object on a temporary schema should not be authorized within a two-phase transaction, as it would cause its state to create dependencies with other sessions, causing all sorts of side effects with the existing session or other sessions spawned later on trying to use the same temporary schema name. Regression tests are added to cover all the grounds found, the original report mentioned function creation, but monitoring closer there are many other patterns with LOCK, DROP or CREATE EXTENSION which are involved. One of the symptoms resulting in combining both is that the session which used the temporary schema is not able to shut down completely, waiting for being able to drop the temporary schema, something that it cannot complete because of the two-phase transaction involved with temporary objects. In this case the client is able to disconnect but the session remains alive on the backend-side, potentially blocking connection backend slots from being used. Other problems reported could also involve server crashes. This is back-patched down to v10, which is where 9b013dc has introduced MyXactFlags, something that this patch relies on. Reported-by: Alexey Bashtanov Author: Michael Paquier Reviewed-by: Masahiko Sawada Discussion: https://postgr.es/m/5d910e2e-0db8-ec06-dd5f-baec420513c3@imap.cc Backpatch-through: 10
2019-01-18 01:21:44 +01:00
if ((MyXactFlags & XACT_FLAGS_ACCESSEDTEMPNAMESPACE))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
Simplify restriction handling of two-phase commit for temporary objects There were two flags used to track the access to temporary tables and to the temporary namespace of a session which are used to restrict PREPARE TRANSACTION, however the first control flag is a concept included in the second. This removes the flag for temporary table tracking, keeping around only the one at namespace level. Author: Michael Paquier Reviewed-by: Álvaro Herrera Discussion: https://postgr.es/m/20190118053126.GH1883@paquier.xyz
2019-01-26 02:45:23 +01:00
errmsg("cannot PREPARE a transaction that has operated on temporary objects")));
Restrict the use of temporary namespace in two-phase transactions Attempting to use a temporary table within a two-phase transaction is forbidden for ages. However, there have been uncovered grounds for a couple of other object types and commands which work on temporary objects with two-phase commit. In short, trying to create, lock or drop an object on a temporary schema should not be authorized within a two-phase transaction, as it would cause its state to create dependencies with other sessions, causing all sorts of side effects with the existing session or other sessions spawned later on trying to use the same temporary schema name. Regression tests are added to cover all the grounds found, the original report mentioned function creation, but monitoring closer there are many other patterns with LOCK, DROP or CREATE EXTENSION which are involved. One of the symptoms resulting in combining both is that the session which used the temporary schema is not able to shut down completely, waiting for being able to drop the temporary schema, something that it cannot complete because of the two-phase transaction involved with temporary objects. In this case the client is able to disconnect but the session remains alive on the backend-side, potentially blocking connection backend slots from being used. Other problems reported could also involve server crashes. This is back-patched down to v10, which is where 9b013dc has introduced MyXactFlags, something that this patch relies on. Reported-by: Alexey Bashtanov Author: Michael Paquier Reviewed-by: Masahiko Sawada Discussion: https://postgr.es/m/5d910e2e-0db8-ec06-dd5f-baec420513c3@imap.cc Backpatch-through: 10
2019-01-18 01:21:44 +01:00
Support synchronization of snapshots through an export/import procedure. A transaction can export a snapshot with pg_export_snapshot(), and then others can import it with SET TRANSACTION SNAPSHOT. The data does not leave the server so there are not security issues. A snapshot can only be imported while the exporting transaction is still running, and there are some other restrictions. I'm not totally convinced that we've covered all the bases for SSI (true serializable) mode, but it works fine for lesser isolation modes. Joachim Wieland, reviewed by Marko Tiikkaja, and rather heavily modified by Tom Lane
2011-10-23 00:22:45 +02:00
/*
* Likewise, don't allow PREPARE after pg_export_snapshot. This could be
* supported if we added cleanup logic to twophase.c, but for now it
* doesn't seem worth the trouble.
*/
if (XactHasExportedSnapshots())
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot PREPARE a transaction that has exported snapshots")));
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* Prevent cancel/die interrupt while cleaning up */
HOLD_INTERRUPTS();
/*
* set the current transaction state information appropriately during
Fix a couple of bogus comments, per Alvaro.
2005-07-14 00:46:09 +02:00
* prepare processing
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*/
s->state = TRANS_PREPARE;
Clean up the rather historically encumbered interface to now() and current time: provide a GetCurrentTimestamp() function that returns current time in the form of a TimestampTz, instead of separate time_t and microseconds fields. This is what all the callers really want anyway, and it eliminates low-level dependencies on AbsoluteTime, which is a deprecated datatype that will have to disappear eventually.
2005-06-30 00:51:57 +02:00
prepared_at = GetCurrentTimestamp();
Add a time-of-preparation column to the pg_prepared_xacts view, per an old suggestion by Oliver Jowett. Also, add a transaction column to the pg_locks view to show the xid of each transaction holding or awaiting locks; this allows prepared transactions to be properly associated with the locks they own. There was already a column named 'transaction', and I chose to rename it to 'transactionid' --- since this column is new in the current devel cycle there should be no backwards compatibility issue to worry about.
2005-06-18 21:33:42 +02:00
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* Reserve the GID for this transaction. This could fail if the requested
* GID is invalid or already in use.
*/
Add a time-of-preparation column to the pg_prepared_xacts view, per an old suggestion by Oliver Jowett. Also, add a transaction column to the pg_locks view to show the xid of each transaction holding or awaiting locks; this allows prepared transactions to be properly associated with the locks they own. There was already a column named 'transaction', and I chose to rename it to 'transactionid' --- since this column is new in the current devel cycle there should be no backwards compatibility issue to worry about.
2005-06-18 21:33:42 +02:00
gxact = MarkAsPreparing(xid, prepareGID, prepared_at,
GetUserId(), MyDatabaseId);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
prepareGID = NULL;
/*
* Collect data for the 2PC state file. Note that in general, no actual
* state change should happen in the called modules during this step,
* since it's still possible to fail before commit, and in that case we
* want transaction abort to be able to clean up. (In particular, the
* AtPrepare routines may error out if they find cases they cannot
* handle.) State cleanup should happen in the PostPrepare routines
* below. However, some modules can go ahead and clear state here because
* they wouldn't do anything with it during abort anyway.
*
* Note: because the 2PC state file records will be replayed in the same
* order they are made, the order of these calls has to match the order in
* which we want things to happen during COMMIT PREPARED or ROLLBACK
* PREPARED; in particular, pay attention to whether things should happen
* before or after releasing the transaction's locks.
*/
StartPrepare(gxact);
AtPrepare_Notify();
AtPrepare_Locks();
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
AtPrepare_PredicateLocks();
Fix up pgstats counting of live and dead tuples to recognize that committed and aborted transactions have different effects; also teach it not to assume that prepared transactions are always committed. Along the way, simplify the pgstats API by tying counting directly to Relations; I cannot detect any redeeming social value in having stats pointers in HeapScanDesc and IndexScanDesc structures. And fix a few corner cases in which counts might be missed because the relation's pgstat_info pointer hadn't been set.
2007-05-27 05:50:39 +02:00
AtPrepare_PgStat();
Fix an old bug in multixact and two-phase commit. Prepared transactions can be part of multixacts, so allocate a slot for each prepared transaction in the "oldest member" array in multixact.c. On PREPARE TRANSACTION, transfer the oldest member value from the current backends slot to the prepared xact slot. Also save and recover the value from the 2pc state file. The symptom of the bug was that after a transaction prepared, a shared lock still held by the prepared transaction was sometimes ignored by other transactions. Fix back to 8.1, where both 2PC and multixact were introduced.
2009-11-23 10:58:36 +01:00
AtPrepare_MultiXact();
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
AtPrepare_RelationMap();
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* Here is where we really truly prepare.
*
* We have to record transaction prepares even if we didn't make any
* updates, because the transaction manager might get confused if we lose
* a global transaction.
*/
EndPrepare(gxact);
/*
* Now we clean up backend-internal state and release internal resources.
*/
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Reset XactLastRecEnd until the next transaction writes something */
Replace XLogRecPtr struct with a 64-bit integer. This simplifies code that needs to do arithmetic on XLogRecPtrs. To avoid changing on-disk format of data pages, the LSN on data pages is still stored in the old format. That should keep pg_upgrade happy. However, we have XLogRecPtrs embedded in the control file, and in the structs that are sent over the replication protocol, so this changes breaks compatibility of pg_basebackup and server. I didn't do anything about this in this patch, per discussion on -hackers, the right thing to do would to be to change the replication protocol to be architecture-independent, so that you could use a newer version of pg_receivexlog, for example, against an older server version.
2012-06-24 17:51:37 +02:00
XactLastRecEnd = 0;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Fix CREATE INDEX CONCURRENTLY for the newest prepared transactions. The purpose of commit 8a54e12a38d1545d249f1402f66c8cde2837d97c was to fix this, and it sufficed when the PREPARE TRANSACTION completed before the CIC looked for lock conflicts. Otherwise, things still broke. As before, in a cluster having used CIC while having enabled prepared transactions, queries that use the resulting index can silently fail to find rows. It may be necessary to reindex to recover from past occurrences; REINDEX CONCURRENTLY suffices. Fix this for future index builds by making CIC wait for arbitrarily-recent prepared transactions and for ordinary transactions that may yet PREPARE TRANSACTION. As part of that, have PREPARE TRANSACTION transfer locks to its dummy PGPROC before it calls ProcArrayClearTransaction(). Back-patch to 9.6 (all supported versions). Andrey Borodin, reviewed (in earlier versions) by Andres Freund. Discussion: https://postgr.es/m/01824242-AA92-4FE9-9BA7-AEBAFFEA3D0C@yandex-team.ru
2021-10-24 03:36:38 +02:00
/*
* Transfer our locks to a dummy PGPROC. This has to be done before
* ProcArrayClearTransaction(). Otherwise, a GetLockConflicts() would
* conclude "xact already committed or aborted" for our locks.
*/
PostPrepare_Locks(xid);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* Let others know about no transaction in progress by me. This has to be
* done *after* the prepared transaction has been marked valid, else
* someone may think it is unlocked and recyclable.
*/
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
ProcArrayClearTransaction(MyProc);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
Fix race condition in preparing a transaction for two-phase commit. To lock a prepared transaction's shared memory entry, we used to mark it with the XID of the backend. When the XID was no longer active according to the proc array, the entry was implicitly considered as not locked anymore. However, when preparing a transaction, the backend's proc array entry was cleared before transfering the locks (and some other state) to the prepared transaction's dummy PGPROC entry, so there was a window where another backend could finish the transaction before it was in fact fully prepared. To fix, rewrite the locking mechanism of global transaction entries. Instead of an XID, just have simple locked-or-not flag in each entry (we store the locking backend's backend id rather than a simple boolean, but that's just for debugging purposes). The backend is responsible for explicitly unlocking the entry, and to make sure that that happens, install a callback to unlock it on abort or process exit. Backpatch to all supported versions.
2014-05-15 15:37:50 +02:00
* In normal commit-processing, this is all non-critical post-transaction
* cleanup. When the transaction is prepared, however, it's important
Fix more typos in comments. Patch by CharSyam, plus a few more I spotted with grep.
2015-05-20 18:44:46 +02:00
* that the locks and other per-backend resources are transferred to the
Fix race condition in preparing a transaction for two-phase commit. To lock a prepared transaction's shared memory entry, we used to mark it with the XID of the backend. When the XID was no longer active according to the proc array, the entry was implicitly considered as not locked anymore. However, when preparing a transaction, the backend's proc array entry was cleared before transfering the locks (and some other state) to the prepared transaction's dummy PGPROC entry, so there was a window where another backend could finish the transaction before it was in fact fully prepared. To fix, rewrite the locking mechanism of global transaction entries. Instead of an XID, just have simple locked-or-not flag in each entry (we store the locking backend's backend id rather than a simple boolean, but that's just for debugging purposes). The backend is responsible for explicitly unlocking the entry, and to make sure that that happens, install a callback to unlock it on abort or process exit. Backpatch to all supported versions.
2014-05-15 15:37:50 +02:00
* prepared transaction's PGPROC entry. Note that if an error is raised
* here, it's too late to abort the transaction. XXX: This probably should
* be in a critical section, to force a PANIC if any of this fails, but
* that cure could be worse than the disease.
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*/
CallXactCallbacks(XACT_EVENT_PREPARE);
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
true, true);
/* Check we've released all buffer pins */
AtEOXact_Buffers(true);
Modify AtEOXact_CatCache and AtEOXact_RelationCache to assume that the ResourceOwner mechanism already released all reference counts for the cache entries; therefore, we do not need to scan the catcache or relcache at transaction end, unless we want to do it as a debugging crosscheck. Do the crosscheck only in Assert mode. This is the same logic we had previously installed in AtEOXact_Buffers to avoid overhead with large numbers of shared buffers. I thought it'd be a good idea to do it here too, in view of Kari Lavikka's recent report showing a real-world case where AtEOXact_CatCache is taking a significant fraction of runtime.
2005-08-08 21:17:23 +02:00
/* Clean up the relation cache */
AtEOXact_RelationCache(true);
Remove flatfiles.c, which is now obsolete. Recent commits have removed the various uses it was supporting. It was a performance bottleneck, according to bug report #4919 by Lauris Ulmanis; seems it slowed down user creation after a billion users.
2009-09-01 04:54:52 +02:00
/* notify doesn't need a postprepare call */
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Fix up pgstats counting of live and dead tuples to recognize that committed and aborted transactions have different effects; also teach it not to assume that prepared transactions are always committed. Along the way, simplify the pgstats API by tying counting directly to Relations; I cannot detect any redeeming social value in having stats pointers in HeapScanDesc and IndexScanDesc structures. And fix a few corner cases in which counts might be missed because the relation's pgstat_info pointer hadn't been set.
2007-05-27 05:50:39 +02:00
PostPrepare_PgStat();
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
PostPrepare_Inval();
PostPrepare_smgr();
Fix an old bug in multixact and two-phase commit. Prepared transactions can be part of multixacts, so allocate a slot for each prepared transaction in the "oldest member" array in multixact.c. On PREPARE TRANSACTION, transfer the oldest member value from the current backends slot to the prepared xact slot. Also save and recover the value from the 2pc state file. The symptom of the bug was that after a transaction prepared, a shared lock still held by the prepared transaction was sometimes ignored by other transactions. Fix back to 8.1, where both 2PC and multixact were introduced.
2009-11-23 10:58:36 +01:00
PostPrepare_MultiXact(xid);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Implement genuine serializable isolation level. Until now, our Serializable mode has in fact been what's called Snapshot Isolation, which allows some anomalies that could not occur in any serialized ordering of the transactions. This patch fixes that using a method called Serializable Snapshot Isolation, based on research papers by Michael J. Cahill (see README-SSI for full references). In Serializable Snapshot Isolation, transactions run like they do in Snapshot Isolation, but a predicate lock manager observes the reads and writes performed and aborts transactions if it detects that an anomaly might occur. This method produces some false positives, ie. it sometimes aborts transactions even though there is no anomaly. To track reads we implement predicate locking, see storage/lmgr/predicate.c. Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared memory is finite, so when a transaction takes many tuple-level locks on a page, the locks are promoted to a single page-level lock, and further to a single relation level lock if necessary. To lock key values with no matching tuple, a sequential scan always takes a relation-level lock, and an index scan acquires a page-level lock that covers the search key, whether or not there are any matching keys at the moment. A predicate lock doesn't conflict with any regular locks or with another predicate locks in the normal sense. They're only used by the predicate lock manager to detect the danger of anomalies. Only serializable transactions participate in predicate locking, so there should be no extra overhead for for other transactions. Predicate locks can't be released at commit, but must be remembered until all the transactions that overlapped with it have completed. That means that we need to remember an unbounded amount of predicate locks, so we apply a lossy but conservative method of tracking locks for committed transactions. If we run short of shared memory, we overflow to a new "pg_serial" SLRU pool. We don't currently allow Serializable transactions in Hot Standby mode. That would be hard, because even read-only transactions can cause anomalies that wouldn't otherwise occur. Serializable isolation mode now means the new fully serializable level. Repeatable Read gives you the old Snapshot Isolation level that we have always had. Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and Anssi Kääriäinen
2011-02-07 22:46:51 +01:00
PostPrepare_PredicateLocks(xid);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_LOCKS,
true, true);
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_AFTER_LOCKS,
true, true);
Fix race condition in preparing a transaction for two-phase commit. To lock a prepared transaction's shared memory entry, we used to mark it with the XID of the backend. When the XID was no longer active according to the proc array, the entry was implicitly considered as not locked anymore. However, when preparing a transaction, the backend's proc array entry was cleared before transfering the locks (and some other state) to the prepared transaction's dummy PGPROC entry, so there was a window where another backend could finish the transaction before it was in fact fully prepared. To fix, rewrite the locking mechanism of global transaction entries. Instead of an XID, just have simple locked-or-not flag in each entry (we store the locking backend's backend id rather than a simple boolean, but that's just for debugging purposes). The backend is responsible for explicitly unlocking the entry, and to make sure that that happens, install a callback to unlock it on abort or process exit. Backpatch to all supported versions.
2014-05-15 15:37:50 +02:00
/*
* Allow another backend to finish the transaction. After
* PostPrepare_Twophase(), the transaction is completely detached from our
* backend. The rest is just non-critical cleanup of backend-local state.
*/
PostPrepare_Twophase();
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* PREPARE acts the same as COMMIT as far as GUC is concerned */
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
AtEOXact_GUC(true, 1);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
AtEOXact_SPI(true);
Relax transactional restrictions on ALTER TYPE ... ADD VALUE (redux). Originally committed as 15bc038f (plus some follow-ups), this was reverted in 28e07270 due to a problem discovered in parallel workers. This new version corrects that problem by sending the list of uncommitted enum values to parallel workers. Here follows the original commit message describing the change: To prevent possibly breaking indexes on enum columns, we must keep uncommitted enum values from getting stored in tables, unless we can be sure that any such column is new in the current transaction. Formerly, we enforced this by disallowing ALTER TYPE ... ADD VALUE from being executed at all in a transaction block, unless the target enum type had been created in the current transaction. This patch removes that restriction, and instead insists that an uncommitted enum value can't be referenced unless it belongs to an enum type created in the same transaction as the value. Per discussion, this should be a bit less onerous. It does require each function that could possibly return a new enum value to SQL operations to check this restriction, but there aren't so many of those that this seems unmaintainable. Author: Andrew Dunstan and Tom Lane, with parallel query fix by Thomas Munro Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAEepm%3D0Ei7g6PaNTbcmAh9tCRahQrk%3Dr5ZWLD-jr7hXweYX3yg%40mail.gmail.com Discussion: https://postgr.es/m/4075.1459088427%40sss.pgh.pa.us
2018-10-09 01:51:01 +02:00
AtEOXact_Enum();
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
AtEOXact_on_commit_actions(true);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
AtEOXact_Namespace(true, false);
Close un-owned SMgrRelations at transaction end. If an SMgrRelation is not "owned" by a relcache entry, don't allow it to live past transaction end. This design allows the same SMgrRelation to be used for blind writes of multiple blocks during a transaction, but ensures that we don't hold onto such an SMgrRelation indefinitely. Because an SMgrRelation typically corresponds to open file descriptors at the fd.c level, leaving it open when there's no corresponding relcache entry can mean that we prevent the kernel from reclaiming deleted disk space. (While CacheInvalidateSmgr messages usually fix that, there are cases where they're not issued, such as DROP DATABASE. We might want to add some more sinval messaging for that, but I'd be inclined to keep this type of logic anyway, since allowing VFDs to accumulate indefinitely for blind-written relations doesn't seem like a good idea.) This code replaces a previous attempt towards the same goal that proved to be unreliable. Back-patch to 9.1 where the previous patch was added.
2012-10-17 18:38:21 +02:00
AtEOXact_SMgr();
In AtEOXact_Files, complain if any files remain unclosed at commit. This change makes this module act more like most of our other low-level resource management modules. It's a caller error if something is not explicitly closed by the end of a successful transaction, so issue a WARNING about it. This would not actually have caught the file leak bug fixed in commit 231bcd080, because that was in a transaction-abort path; but it still seems like a good, and pretty cheap, cross-check. Discussion: https://postgr.es/m/152056616579.4966.583293218357089052@wrigleys.postgresql.org
2018-04-28 23:45:02 +02:00
AtEOXact_Files(true);
Combine cmin and cmax fields of HeapTupleHeaders into a single field, by keeping private state in each backend that has inserted and deleted the same tuple during its current top-level transaction. This is sufficient since there is no need to be able to determine the cmin/cmax from any other transaction. This gets us back down to 23-byte headers, removing a penalty paid in 8.0 to support subtransactions. Patch by Heikki Linnakangas, with minor revisions by moi, following a design hashed out awhile back on the pghackers list.
2007-02-09 04:35:35 +01:00
AtEOXact_ComboCid();
Fix dynahash.c to suppress hash bucket splits while a hash_seq_search() scan is in progress on the same hashtable. This seems the least invasive way to fix the recently-recognized problem that a split could cause the scan to visit entries twice or (with much lower probability) miss them entirely. The only field-reported problem caused by this is the "failed to re-find shared lock object" PANIC in COMMIT PREPARED reported by Michel Dorochevsky, which was caused by multiply visited entries. However, it seems certain that mdsync() is vulnerable to missing required fsync's due to missed entries, and I am fearful that RelationCacheInitializePhase2() might be at risk as well. Because of that and the generalized hazard presented by this bug, back-patch all the supported branches. Along the way, fix pg_prepared_statement() and pg_cursor() to not assume that the hashtables they are examining will stay static between calls. This is risky regardless of the newly noted dynahash problem, because hash_seq_search() has never promised to cope with deletion of table entries other than the just-returned one. There may be no bug here because the only supported way to call these functions is via ExecMakeTableFunctionResult() which will cycle them to completion before doing anything very interesting, but it seems best to get rid of the assumption. This affects 8.2 and HEAD only, since those functions weren't there earlier.
2007-04-27 01:24:46 +02:00
AtEOXact_HashTables(true);
Reset pg_stat_activity.xact_start during PREPARE TRANSACTION. Once we've completed a PREPARE, our session is not running a transaction, so its entry in pg_stat_activity should show xact_start as null, rather than leaving the value as the start time of the now-prepared transaction. I think possibly this oversight was triggered by faulty extrapolation from the adjacent comment that says PrepareTransaction should not call AtEOXact_PgStat, so tweak the wording of that comment. Noted by Andres Freund while considering bug #10123 from Maxim Boguk, although this error doesn't seem to explain that report. Back-patch to all active branches.
2014-04-24 19:29:48 +02:00
/* don't call AtEOXact_PgStat here; we fixed pgstat state above */
Avoid SnapshotResetXmin() during AtEOXact_Snapshot() For normal commits and aborts we already reset PgXact->xmin, so we can simply avoid running SnapshotResetXmin() twice. During performance tests by Alexander Korotkov, diagnosis by Andres Freund showed PgXact array as a bottleneck. After manual analysis by me of the code paths that touch those memory locations, I was able to identify extraneous code in the main transaction commit path. Avoiding touching highly contented shmem improves concurrent performance slightly on all workloads, confirmed by tests run by Ashutosh Sharma and Alexander Korotkov. Simon Riggs Discussion: CANP8+jJdXE9b+b9F8CQT-LuxxO0PBCB-SZFfMVAdp+akqo4zfg@mail.gmail.com
2017-04-06 14:31:52 +02:00
AtEOXact_Snapshot(true, true);
Wake up a subscription's replication worker processes after DDL. Waken related worker processes immediately at commit of a transaction that has performed ALTER SUBSCRIPTION (including the RENAME and OWNER variants). This reduces the response time for such operations. In the real world that might not be worth much, but it shaves several seconds off the runtime for the subscription test suite. In the case of PREPARE, we just throw away this notification state; it doesn't seem worth the work to preserve it. The workers will still react after the eventual COMMIT PREPARED, but not as quickly. Nathan Bossart Discussion: https://postgr.es/m/20221122004119.GA132961@nathanxps13
2023-01-06 22:08:20 +01:00
/* we treat PREPARE as ROLLBACK so far as waking workers goes */
AtEOXact_ApplyLauncher(false);
AtEOXact_LogicalRepWorkers(false);
Reset pg_stat_activity.xact_start during PREPARE TRANSACTION. Once we've completed a PREPARE, our session is not running a transaction, so its entry in pg_stat_activity should show xact_start as null, rather than leaving the value as the start time of the now-prepared transaction. I think possibly this oversight was triggered by faulty extrapolation from the adjacent comment that says PrepareTransaction should not call AtEOXact_PgStat, so tweak the wording of that comment. Noted by Andres Freund while considering bug #10123 from Maxim Boguk, although this error doesn't seem to explain that report. Back-patch to all active branches.
2014-04-24 19:29:48 +02:00
pgstat_report_xact_timestamp(0);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
CurrentResourceOwner = NULL;
ResourceOwnerDelete(TopTransactionResourceOwner);
s->curTransactionOwner = NULL;
CurTransactionResourceOwner = NULL;
TopTransactionResourceOwner = NULL;
AtCommit_Memory();
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = InvalidFullTransactionId;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
s->subTransactionId = InvalidSubTransactionId;
s->nestingLevel = 0;
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
s->gucNestLevel = 0;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
XactTopFullTransactionId = InvalidFullTransactionId;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nParallelCurrentXids = 0;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* done with 1st phase commit processing, set current transaction state
* back to default
*/
s->state = TRANS_DEFAULT;
RESUME_INTERRUPTS();
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AbortTransaction
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AbortTransaction(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
TransactionId latestXid;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
bool is_parallel_worker;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Restructure backend SIGINT/SIGTERM handling so that 'die' interrupts are treated more like 'cancel' interrupts: the signal handler sets a flag that is examined at well-defined spots, rather than trying to cope with an interrupt that might happen anywhere. See pghackers discussion of 1/12/01.
2001-01-14 06:08:17 +01:00
/* Prevent cancel/die interrupt while cleaning up */
Make critical sections (elog->crash) and interrupt holdoff sections into distinct concepts, per recent discussion on pghackers.
2001-01-19 23:08:47 +01:00
HOLD_INTERRUPTS();
Restructure backend SIGINT/SIGTERM handling so that 'die' interrupts are treated more like 'cancel' interrupts: the signal handler sets a flag that is examined at well-defined spots, rather than trying to cope with an interrupt that might happen anywhere. See pghackers discussion of 1/12/01.
2001-01-14 06:08:17 +01:00
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/* Make sure we have a valid memory context and resource owner */
AtAbort_Memory();
AtAbort_ResourceOwner();
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
/*
Implement new 'lightweight lock manager' that's intermediate between existing lock manager and spinlocks: it understands exclusive vs shared lock but has few other fancy features. Replace most uses of spinlocks with lightweight locks. All remaining uses of spinlocks have very short lock hold times (a few dozen instructions), so tweak spinlock backoff code to work efficiently given this assumption. All per my proposal on pghackers 26-Sep-01.
2001-09-29 06:02:27 +02:00
* Release any LW locks we might be holding as quickly as possible.
* (Regular locks, however, must be held till we finish aborting.)
* Releasing LW locks is critical since we might try to grab them again
* while cleaning up!
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
*/
Implement new 'lightweight lock manager' that's intermediate between existing lock manager and spinlocks: it understands exclusive vs shared lock but has few other fancy features. Replace most uses of spinlocks with lightweight locks. All remaining uses of spinlocks have very short lock hold times (a few dozen instructions), so tweak spinlock backoff code to work efficiently given this assumption. All per my proposal on pghackers 26-Sep-01.
2001-09-29 06:02:27 +02:00
LWLockReleaseAll();
Provide much better wait information in pg_stat_activity. When a process is waiting for a heavyweight lock, we will now indicate the type of heavyweight lock for which it is waiting. Also, you can now see when a process is waiting for a lightweight lock - in which case we will indicate the individual lock name or the tranche, as appropriate - or for a buffer pin. Amit Kapila, Ildus Kurbangaliev, reviewed by me. Lots of helpful discussion and suggestions by many others, including Alexander Korotkov, Vladimir Borodin, and many others.
2016-03-10 18:44:09 +01:00
/* Clear wait information and command progress indicator */
pgstat_report_wait_end();
pgstat_progress_end_command();
bufmgr: Support multiple in-progress IOs by using resowner A future patch will add support for extending relations by multiple blocks at once. To be concurrency safe, the buffers for those blocks need to be marked as BM_IO_IN_PROGRESS. Until now we only had infrastructure for recovering from an IO error for a single buffer. This commit extends that infrastructure to multiple buffers by using the resource owner infrastructure. This commit increases the size of the ResourceOwnerData struct, which appears to have a just about measurable overhead in very extreme workloads. Medium term we are planning to substantially shrink the size of ResourceOwnerData. Short term the increase is small enough to not worry about it for now. Reviewed-by: Melanie Plageman <melanieplageman@gmail.com> Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de Discussion: https://postgr.es/m/20221029200025.w7bvlgvamjfo6z44@awork3.anarazel.de
2023-04-05 22:55:15 +02:00
/* Clean up buffer context locks, too */
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
UnlockBuffers();
pgindent run. Make it all clean.
2001-03-22 05:01:46 +01:00
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
/* Reset WAL record construction state */
XLogResetInsertion();
Support condition variables. Condition variables provide a flexible way to sleep until a cooperating process causes an arbitrary condition to become true. In simple cases, this can be accomplished with a WaitLatch/ResetLatch loop; the cooperating process can call SetLatch after performing work that might cause the condition to be satisfied, and the waiting process can recheck the condition each time. However, if the process performing the work doesn't have an easy way to identify which processes might be waiting, this doesn't work, because it can't identify which latches to set. Condition variables solve that problem by internally maintaining a list of waiters; a process that may have caused some waiter's condition to be satisfied must "signal" or "broadcast" on the condition variable. Robert Haas and Thomas Munro
2016-11-22 20:26:40 +01:00
/* Cancel condition variable sleep */
ConditionVariableCancelSleep();
Restructure backend SIGINT/SIGTERM handling so that 'die' interrupts are treated more like 'cancel' interrupts: the signal handler sets a flag that is examined at well-defined spots, rather than trying to cope with an interrupt that might happen anywhere. See pghackers discussion of 1/12/01.
2001-01-14 06:08:17 +01:00
/*
* Also clean up any open wait for lock, since the lock manager will choke
* if we try to wait for another lock before doing this.
*/
Tighten up error recovery for fast-path locking. The previous code could cause a backend crash after BEGIN; SAVEPOINT a; LOCK TABLE foo (interrupted by ^C or statement timeout); ROLLBACK TO SAVEPOINT a; LOCK TABLE foo, and might have leaked strong-lock counts in other situations. Report by Zoltán Böszörményi; patch review by Jeff Davis.
2012-04-18 17:17:30 +02:00
LockErrorCleanup();
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
/*
* If any timeout events are still active, make sure the timeout interrupt
* is scheduled. This covers possible loss of a timeout interrupt due to
* longjmp'ing out of the SIGINT handler (see notes in handle_sig_alarm).
* We delay this till after LockErrorCleanup so that we don't uselessly
* reschedule lock or deadlock check timeouts.
*/
reschedule_timeouts();
/*
* Re-enable signals, in case we got here by longjmp'ing out of a signal
* handler. We do this fairly early in the sequence so that the timeout
* infrastructure will be functional if needed while aborting.
*/
Retire PG_SETMASK() macro. In the 90s we needed to deal with computers that still had the pre-standard signal masking APIs. That hasn't been relevant for a very long time on Unix systems, and c94ae9d8 got rid of a remaining dependency in our Windows porting code. PG_SETMASK didn't expose save/restore functionality, so we'd already started using sigprocmask() directly in places, creating the visual distraction of having two ways to spell it. It's not part of the API that extensions are expected to be using (but if they are, the change will be trivial). It seems like a good time to drop the old macro and just call the standard POSIX function. Reviewed-by: Nathan Bossart <nathandbossart@gmail.com> Discussion: https://postgr.es/m/CA%2BhUKG%2BKfQgrhHP2DLTohX1WwubaCBHmTzGnAEDPZ-Gug-Xskg%40mail.gmail.com
2023-02-02 22:34:56 +01:00
sigprocmask(SIG_SETMASK, &UnBlockSig, NULL);
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* check the current transaction state
*/
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
is_parallel_worker = (s->blockState == TBLOCK_PARALLEL_INPROGRESS);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
if (s->state != TRANS_INPROGRESS && s->state != TRANS_PREPARE)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "AbortTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Assert(s->parent == NULL);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* set the current transaction state information appropriately during the
* abort processing
*/
s->state = TRANS_ABORT;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
/*
Make standard maintenance operations (including VACUUM, ANALYZE, REINDEX, and CLUSTER) execute as the table owner rather than the calling user, using the same privilege-switching mechanism already used for SECURITY DEFINER functions. The purpose of this change is to ensure that user-defined functions used in index definitions cannot acquire the privileges of a superuser account that is performing routine maintenance. While a function used in an index is supposed to be IMMUTABLE and thus not able to do anything very interesting, there are several easy ways around that restriction; and even if we could plug them all, there would remain a risk of reading sensitive information and broadcasting it through a covert channel such as CPU usage. To prevent bypassing this security measure, execution of SET SESSION AUTHORIZATION and SET ROLE is now forbidden within a SECURITY DEFINER context. Thanks to Itagaki Takahiro for reporting this vulnerability. Security: CVE-2007-6600
2008-01-03 22:23:15 +01:00
* Reset user ID which might have been changed transiently. We need this
* to clean up in case control escaped out of a SECURITY DEFINER function
* or other local change of CurrentUserId; therefore, the prior value of
Prevent indirect security attacks via changing session-local state within an allegedly immutable index function. It was previously recognized that we had to prevent such a function from executing SET/RESET ROLE/SESSION AUTHORIZATION, or it could trivially obtain the privileges of the session user. However, since there is in general no privilege checking for changes of session-local state, it is also possible for such a function to change settings in a way that might subvert later operations in the same session. Examples include changing search_path to cause an unexpected function to be called, or replacing an existing prepared statement with another one that will execute a function of the attacker's choosing. The present patch secures VACUUM, ANALYZE, and CREATE INDEX/REINDEX against these threats, which are the same places previously deemed to need protection against the SET ROLE issue. GUC changes are still allowed, since there are many useful cases for that, but we prevent security problems by forcing a rollback of any GUC change after completing the operation. Other cases are handled by throwing an error if any change is attempted; these include temp table creation, closing a cursor, and creating or deleting a prepared statement. (In 7.4, the infrastructure to roll back GUC changes doesn't exist, so we settle for rejecting changes of "search_path" in these contexts.) Original report and patch by Gurjeet Singh, additional analysis by Tom Lane. Security: CVE-2009-4136
2009-12-09 22:57:51 +01:00
* SecurityRestrictionContext also needs to be restored.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*
Make standard maintenance operations (including VACUUM, ANALYZE, REINDEX, and CLUSTER) execute as the table owner rather than the calling user, using the same privilege-switching mechanism already used for SECURITY DEFINER functions. The purpose of this change is to ensure that user-defined functions used in index definitions cannot acquire the privileges of a superuser account that is performing routine maintenance. While a function used in an index is supposed to be IMMUTABLE and thus not able to do anything very interesting, there are several easy ways around that restriction; and even if we could plug them all, there would remain a risk of reading sensitive information and broadcasting it through a covert channel such as CPU usage. To prevent bypassing this security measure, execution of SET SESSION AUTHORIZATION and SET ROLE is now forbidden within a SECURITY DEFINER context. Thanks to Itagaki Takahiro for reporting this vulnerability. Security: CVE-2007-6600
2008-01-03 22:23:15 +01:00
* (Note: it is not necessary to restore session authorization or role
* settings here because those can only be changed via GUC, and GUC will
* take care of rolling them back if need be.)
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
*/
Prevent indirect security attacks via changing session-local state within an allegedly immutable index function. It was previously recognized that we had to prevent such a function from executing SET/RESET ROLE/SESSION AUTHORIZATION, or it could trivially obtain the privileges of the session user. However, since there is in general no privilege checking for changes of session-local state, it is also possible for such a function to change settings in a way that might subvert later operations in the same session. Examples include changing search_path to cause an unexpected function to be called, or replacing an existing prepared statement with another one that will execute a function of the attacker's choosing. The present patch secures VACUUM, ANALYZE, and CREATE INDEX/REINDEX against these threats, which are the same places previously deemed to need protection against the SET ROLE issue. GUC changes are still allowed, since there are many useful cases for that, but we prevent security problems by forcing a rollback of any GUC change after completing the operation. Other cases are handled by throwing an error if any change is attempted; these include temp table creation, closing a cursor, and creating or deleting a prepared statement. (In 7.4, the infrastructure to roll back GUC changes doesn't exist, so we settle for rejecting changes of "search_path" in these contexts.) Original report and patch by Gurjeet Singh, additional analysis by Tom Lane. Security: CVE-2009-4136
2009-12-09 22:57:51 +01:00
SetUserIdAndSecContext(s->prevUser, s->prevSecContext);
Clean up backend-exit-time cleanup behavior. Use on_shmem_exit callbacks to ensure that we have released buffer refcounts and so forth, rather than putting ad-hoc operations before (some of the calls to) proc_exit. Add commentary to discourage future hackers from repeating that mistake.
2000-12-18 01:44:50 +01:00
Fix possible crash during FATAL exit from reindexing. index.c supposed that it could just use a PG_TRY block to clean up the state associated with an active REINDEX operation. However, that code doesn't run if we do a FATAL exit --- for example, due to a SIGTERM shutdown signal --- while the REINDEX is happening. And that state does get consulted during catalog accesses, which makes it problematic if we do any catalog accesses during shutdown --- for example, to clean up any temp tables created in the session. If this combination of circumstances occurred, we could find ourselves trying to access already-freed memory. In debug builds that'd fairly reliably cause an assertion failure. In production we might often get away with it, but with some bad luck it could cause a core dump. Another possible bad outcome is an erroneous conclusion that an index-to-be-accessed is being reindexed; but it looks like that would be unlikely to have any consequences worse than failing to drop temp tables right away. (They'd still get dropped by the next session that uses that temp schema.) To fix, get rid of the use of PG_TRY here, and instead hook into the transaction abort mechanisms to clean up reindex state. Per bug #16378 from Alexander Lakhin. This has been wrong for a very long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/16378-7a70ca41b3ec2009@postgresql.org
2020-04-21 21:58:42 +02:00
/* Forget about any active REINDEX. */
ResetReindexState(s->nestingLevel);
Implement streaming mode in ReorderBuffer. Instead of serializing the transaction to disk after reaching the logical_decoding_work_mem limit in memory, we consume the changes we have in memory and invoke stream API methods added by commit 45fdc9738b. However, sometimes if we have incomplete toast or speculative insert we spill to the disk because we can't generate the complete tuple and stream. And, as soon as we get the complete tuple we stream the transaction including the serialized changes. We can do this incremental processing thanks to having assignments (associating subxact with toplevel xacts) in WAL right away, and thanks to logging the invalidation messages at each command end. These features are added by commits 0bead9af48 and c55040ccd0 respectively. Now that we can stream in-progress transactions, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We handle such failures by returning ERRCODE_TRANSACTION_ROLLBACK sqlerrcode from system table scan APIs to the backend or WALSender decoding a specific uncommitted transaction. The decoding logic on the receipt of such a sqlerrcode aborts the decoding of the current transaction and continue with the decoding of other transactions. We have ReorderBufferTXN pointer in each ReorderBufferChange by which we know which xact it belongs to. The output plugin can use this to decide which changes to discard in case of stream_abort_cb (e.g. when a subxact gets discarded). We also provide a new option via SQL APIs to fetch the changes being streamed. Author: Dilip Kumar, Tomas Vondra, Amit Kapila, Nikhil Sontakke Reviewed-by: Amit Kapila, Kuntal Ghosh, Ajin Cherian Tested-by: Neha Sharma, Mahendra Singh Thalor and Ajin Cherian Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2020-08-08 04:04:39 +02:00
/* Reset logical streaming state. */
ResetLogicalStreamingState();
Reset properly snapshot export state during transaction abort During a replication slot creation, an ERROR generated in the same transaction as the one creating a to-be-exported snapshot would have left the backend in an inconsistent state, as the associated static export snapshot state was not being reset on transaction abort, but only on the follow-up command received by the WAL sender that created this snapshot on replication slot creation. This would trigger inconsistency failures if this session tried to export again a snapshot, like during the creation of a replication slot. Note that a snapshot export cannot happen in a transaction block, so there is no need to worry resetting this state for subtransaction aborts. Also, this inconsistent state would very unlikely show up to users. For example, one case where this could happen is an out-of-memory error when building the initial snapshot to-be-exported. Dilip found this problem while poking at a different patch, that caused an error in this code path for reasons unrelated to HEAD. Author: Dilip Kumar Reviewed-by: Michael Paquier, Zhihong Yu Discussion: https://postgr.es/m/CAFiTN-s0zA1Kj0ozGHwkYkHwa5U0zUE94RSc_g81WrpcETB5=w@mail.gmail.com Backpatch-through: 9.6
2021-10-18 04:55:42 +02:00
/* Reset snapshot export state. */
SnapBuildResetExportedSnapshotState();
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/* If in parallel mode, clean up workers and exit parallel mode. */
if (IsInParallelMode())
{
AtEOXact_Parallel(false);
s->parallelModeLevel = 0;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* do abort processing
*/
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
AfterTriggerEndXact(false); /* 'false' means it's abort */
Portal and memory management infrastructure for extended query protocol. Both plannable queries and utility commands are now always executed within Portals, which have been revamped so that they can handle the load (they used to be good only for single SELECT queries). Restructure code to push command-completion-tag selection logic out of postgres.c, so that it won't have to be duplicated between simple and extended queries. initdb forced due to addition of a field to Query nodes.
2003-05-02 22:54:36 +02:00
AtAbort_Portals();
Skip WAL for new relfilenodes, under wal_level=minimal. Until now, only selected bulk operations (e.g. COPY) did this. If a given relfilenode received both a WAL-skipping COPY and a WAL-logged operation (e.g. INSERT), recovery could lose tuples from the COPY. See src/backend/access/transam/README section "Skipping WAL for New RelFileNode" for the new coding rules. Maintainers of table access methods should examine that section. To maintain data durability, just before commit, we choose between an fsync of the relfilenode and copying its contents to WAL. A new GUC, wal_skip_threshold, guides that choice. If this change slows a workload that creates small, permanent relfilenodes under wal_level=minimal, try adjusting wal_skip_threshold. Users setting a timeout on COMMIT may need to adjust that timeout, and log_min_duration_statement analysis will reflect time consumption moving to COMMIT from commands like COPY. Internally, this requires a reliable determination of whether RollbackAndReleaseCurrentSubTransaction() would unlink a relation's current relfilenode. Introduce rd_firstRelfilenodeSubid. Amend the specification of rd_createSubid such that the field is zero when a new rel has an old rd_node. Make relcache.c retain entries for certain dropped relations until end of transaction. Bump XLOG_PAGE_MAGIC, since this introduces XLOG_GIST_ASSIGN_LSN. Future servers accept older WAL, so this bump is discretionary. Kyotaro Horiguchi, reviewed (in earlier, similar versions) by Robert Haas. Heikki Linnakangas and Michael Paquier implemented earlier designs that materially clarified the problem. Reviewed, in earlier designs, by Andrew Dunstan, Andres Freund, Alvaro Herrera, Tom Lane, Fujii Masao, and Simon Riggs. Reported by Martijn van Oosterhout. Discussion: https://postgr.es/m/20150702220524.GA9392@svana.org
2020-04-04 21:25:34 +02:00
smgrDoPendingSyncs(false, is_parallel_worker);
Create a "relation mapping" infrastructure to support changing the relfilenodes of shared or nailed system catalogs. This has two key benefits: * The new CLUSTER-based VACUUM FULL can be applied safely to all catalogs. * We no longer have to use an unsafe reindex-in-place approach for reindexing shared catalogs. CLUSTER on nailed catalogs now works too, although I left it disabled on shared catalogs because the resulting pg_index.indisclustered update would only be visible in one database. Since reindexing shared system catalogs is now fully transactional and crash-safe, the former special cases in REINDEX behavior have been removed; shared catalogs are treated the same as non-shared. This commit does not do anything about the recently-discussed problem of deadlocks between VACUUM FULL/CLUSTER on a system catalog and other concurrent queries; will address that in a separate patch. As a stopgap, parallel_schedule has been tweaked to run vacuum.sql by itself, to avoid such failures during the regression tests.
2010-02-07 21:48:13 +01:00
AtEOXact_LargeObject(false);
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
AtAbort_Notify();
Handle parallel index builds on mapped relations. Commit 9da0cc35284, which introduced parallel CREATE INDEX, failed to propagate relmapper.c backend local cache state to parallel worker processes. This could result in parallel index builds against mapped catalog relations where the leader process (participating as a worker) scans the new, pristine relfilenode, while worker processes scan the obsolescent relfilenode. When this happened, the final index structure was typically not consistent with the owning table's structure. The final index structure could contain entries formed from both heap relfilenodes. Only rebuilds on mapped catalog relations that occur as part of a VACUUM FULL or CLUSTER could become corrupt in practice, since their mapped relation relfilenode swap is what allows the inconsistency to arise. On master, fix the problem by propagating the required relmapper.c backend state as part of standard parallel initialization (Cf. commit 29d58fd3). On v11, simply disallow builds against mapped catalog relations by deeming them parallel unsafe. Author: Peter Geoghegan Reported-By: "death lock" Reviewed-By: Tom Lane, Amit Kapila Bug: #15309 Discussion: https://postgr.es/m/153329671686.1405.18298309097348420351@wrigleys.postgresql.org Backpatch: 11-, where parallel CREATE INDEX was introduced.
2018-08-10 22:01:34 +02:00
AtEOXact_RelationMap(false, is_parallel_worker);
Fix race condition in preparing a transaction for two-phase commit. To lock a prepared transaction's shared memory entry, we used to mark it with the XID of the backend. When the XID was no longer active according to the proc array, the entry was implicitly considered as not locked anymore. However, when preparing a transaction, the backend's proc array entry was cleared before transfering the locks (and some other state) to the prepared transaction's dummy PGPROC entry, so there was a window where another backend could finish the transaction before it was in fact fully prepared. To fix, rewrite the locking mechanism of global transaction entries. Instead of an XID, just have simple locked-or-not flag in each entry (we store the locking backend's backend id rather than a simple boolean, but that's just for debugging purposes). The backend is responsible for explicitly unlocking the entry, and to make sure that that happens, install a callback to unlock it on abort or process exit. Backpatch to all supported versions.
2014-05-15 15:37:50 +02:00
AtAbort_Twophase();
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
Add code to prevent transaction ID wraparound by enforcing a safe limit in GetNewTransactionId(). Since the limit value has to be computed before we run any real transactions, this requires adding code to database startup to scan pg_database and determine the oldest datfrozenxid. This can conveniently be combined with the first stage of an attack on the problem that the 'flat file' copies of pg_shadow and pg_group are not properly updated during WAL recovery. The code I've added to startup resides in a new file src/backend/utils/init/flatfiles.c, and it is responsible for rewriting the flat files as well as initializing the XID wraparound limit value. This will eventually allow us to get rid of GetRawDatabaseInfo too, but we'll need an initdb so we can add a trigger to pg_database.
2005-02-20 03:22:07 +01:00
/*
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
* Advertise the fact that we aborted in pg_xact (assuming that we got as
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* far as assigning an XID to advertise). But if we're inside a parallel
* worker, skip this; the user backend must be the one to write the abort
* record.
Add code to prevent transaction ID wraparound by enforcing a safe limit in GetNewTransactionId(). Since the limit value has to be computed before we run any real transactions, this requires adding code to database startup to scan pg_database and determine the oldest datfrozenxid. This can conveniently be combined with the first stage of an attack on the problem that the 'flat file' copies of pg_shadow and pg_group are not properly updated during WAL recovery. The code I've added to startup resides in a new file src/backend/utils/init/flatfiles.c, and it is responsible for rewriting the flat files as well as initializing the XID wraparound limit value. This will eventually allow us to get rid of GetRawDatabaseInfo too, but we'll need an initdb so we can add a trigger to pg_database.
2005-02-20 03:22:07 +01:00
*/
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
if (!is_parallel_worker)
latestXid = RecordTransactionAbort(false);
else
{
latestXid = InvalidTransactionId;
/*
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* Since the parallel leader won't get our value of XactLastRecEnd in
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* this case, we nudge WAL-writer ourselves in this case. See related
* comments in RecordTransactionAbort for why this matters.
*/
XLogSetAsyncXactLSN(XactLastRecEnd);
}
Make DROP TABLE rollback-able: postpone physical file delete until commit. (WAL logging for this is not done yet, however.) Clean up a number of really crufty things that are no longer needed now that DROP behaves nicely. Make temp table mapper do the right things when drop or rename affecting a temp table is rolled back. Also, remove "relation modified while in use" error check, in favor of locking tables at first reference and holding that lock throughout the statement.
2000-11-08 23:10:03 +01:00
Enable probes to work with Mac OS X Leopard and other OSes that will support DTrace in the future. Switch from using DTRACE_PROBEn macros to the dynamically generated macros. Use "dtrace -h" to create a header file that contains the dynamically generated macros to be used in the source code instead of the DTRACE_PROBEn macros. A dummy header file is generated for builds without DTrace support. Author: Robert Lor <Robert.Lor@sun.com>
2008-03-17 20:44:41 +01:00
TRACE_POSTGRESQL_TRANSACTION_ABORT(MyProc->lxid);
Don't take ProcArrayLock while exiting a transaction that has no XID; there is no need for serialization against snapshot-taking because the xact doesn't affect anyone else's snapshot anyway. Per discussion. Also, move various info about the interlocking of transactions and snapshots out of code comments and into a hopefully-more-cohesive discussion in access/transam/README. Also, remove a couple of now-obsolete comments about having to force some WAL to be written to persuade RecordTransactionCommit to do its thing.
2007-09-07 22:59:26 +02:00
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
/*
* Let others know about no transaction in progress by me. Note that this
* must be done _before_ releasing locks we hold and _after_
* RecordTransactionAbort.
*/
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
ProcArrayEndTransaction(MyProc, latestXid);
Statistical system views (yet without the config stuff, but it's hard to keep such massive changes in sync with the tree so I need to get it in and work from there now). Jan
2001-06-22 21:16:24 +02:00
Perform transaction cleanup operations in a less ad-hoc, more principled order; in particular ensure that all shared resources are released before we release transaction locks. The code used to release locks before buffer pins, which might explain an ancient note I have about a bufmgr assertion failure I'd seen once several years ago, and been unable to reproduce since. (Theory: someone trying to drop a relation might be able to reach FlushRelationBuffers before the last user of the relation had gotten around to dropping his buffer pins.)
2002-10-23 00:44:36 +02:00
/*
* Post-abort cleanup. See notes in CommitTransaction() concerning
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
* ordering. We can skip all of it if the transaction failed before
* creating a resource owner.
Perform transaction cleanup operations in a less ad-hoc, more principled order; in particular ensure that all shared resources are released before we release transaction locks. The code used to release locks before buffer pins, which might explain an ancient note I have about a bufmgr assertion failure I'd seen once several years ago, and been unable to reproduce since. (Theory: someone trying to drop a relation might be able to reach FlushRelationBuffers before the last user of the relation had gotten around to dropping his buffer pins.)
2002-10-23 00:44:36 +02:00
*/
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
if (TopTransactionResourceOwner != NULL)
{
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
if (is_parallel_worker)
CallXactCallbacks(XACT_EVENT_PARALLEL_ABORT);
else
CallXactCallbacks(XACT_EVENT_ABORT);
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
false, true);
AtEOXact_Buffers(false);
AtEOXact_RelationCache(false);
AtEOXact_Inval(false);
AtEOXact_MultiXact();
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_LOCKS,
false, true);
ResourceOwnerRelease(TopTransactionResourceOwner,
RESOURCE_RELEASE_AFTER_LOCKS,
false, true);
During transaction cleanup, release locks before deleting files. There's no need to hold onto the locks until the files are needed, and by doing it this way, we reduce the impact on other backends who may be awaiting locks we hold. Noah Misch
2012-06-14 16:19:33 +02:00
smgrDoPendingDeletes(false);
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
AtEOXact_GUC(false, 1);
AtEOXact_SPI(false);
Relax transactional restrictions on ALTER TYPE ... ADD VALUE (redux). Originally committed as 15bc038f (plus some follow-ups), this was reverted in 28e07270 due to a problem discovered in parallel workers. This new version corrects that problem by sending the list of uncommitted enum values to parallel workers. Here follows the original commit message describing the change: To prevent possibly breaking indexes on enum columns, we must keep uncommitted enum values from getting stored in tables, unless we can be sure that any such column is new in the current transaction. Formerly, we enforced this by disallowing ALTER TYPE ... ADD VALUE from being executed at all in a transaction block, unless the target enum type had been created in the current transaction. This patch removes that restriction, and instead insists that an uncommitted enum value can't be referenced unless it belongs to an enum type created in the same transaction as the value. Per discussion, this should be a bit less onerous. It does require each function that could possibly return a new enum value to SQL operations to check this restriction, but there aren't so many of those that this seems unmaintainable. Author: Andrew Dunstan and Tom Lane, with parallel query fix by Thomas Munro Reviewed-by: Tom Lane Discussion: https://postgr.es/m/CAEepm%3D0Ei7g6PaNTbcmAh9tCRahQrk%3Dr5ZWLD-jr7hXweYX3yg%40mail.gmail.com Discussion: https://postgr.es/m/4075.1459088427%40sss.pgh.pa.us
2018-10-09 01:51:01 +02:00
AtEOXact_Enum();
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
AtEOXact_on_commit_actions(false);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
AtEOXact_Namespace(false, is_parallel_worker);
Close un-owned SMgrRelations at transaction end. If an SMgrRelation is not "owned" by a relcache entry, don't allow it to live past transaction end. This design allows the same SMgrRelation to be used for blind writes of multiple blocks during a transaction, but ensures that we don't hold onto such an SMgrRelation indefinitely. Because an SMgrRelation typically corresponds to open file descriptors at the fd.c level, leaving it open when there's no corresponding relcache entry can mean that we prevent the kernel from reclaiming deleted disk space. (While CacheInvalidateSmgr messages usually fix that, there are cases where they're not issued, such as DROP DATABASE. We might want to add some more sinval messaging for that, but I'd be inclined to keep this type of logic anyway, since allowing VFDs to accumulate indefinitely for blind-written relations doesn't seem like a good idea.) This code replaces a previous attempt towards the same goal that proved to be unreliable. Back-patch to 9.1 where the previous patch was added.
2012-10-17 18:38:21 +02:00
AtEOXact_SMgr();
In AtEOXact_Files, complain if any files remain unclosed at commit. This change makes this module act more like most of our other low-level resource management modules. It's a caller error if something is not explicitly closed by the end of a successful transaction, so issue a WARNING about it. This would not actually have caught the file leak bug fixed in commit 231bcd080, because that was in a transaction-abort path; but it still seems like a good, and pretty cheap, cross-check. Discussion: https://postgr.es/m/152056616579.4966.583293218357089052@wrigleys.postgresql.org
2018-04-28 23:45:02 +02:00
AtEOXact_Files(false);
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
AtEOXact_ComboCid();
AtEOXact_HashTables(false);
Avoid counting transaction stats for parallel worker cooperating transaction. The transaction that is initiated by the parallel worker to cooperate with the actual transaction started by the main backend to complete the query execution should not be counted as a separate transaction. The other internal transactions started and committed by the parallel worker are still counted as separate transactions as we that is what we do in other places like autovacuum. This will partially fix the bloat in transaction stats due to additional transactions performed by parallel workers. For a complete fix, we need to decide how we want to show all the transactions that are started internally for various operations and that is a matter of separate patch. Reported-by: Haribabu Kommi Author: Haribabu Kommi Reviewed-by: Amit Kapila, Jamison Kirk and Rahila Syed Backpatch-through: 9.6 Discussion: https://postgr.es/m/CAJrrPGc9=jKXuScvNyQ+VNhO0FZk7LLAShAJRyZjnedd2D61EQ@mail.gmail.com
2019-04-10 04:54:15 +02:00
AtEOXact_PgStat(false, is_parallel_worker);
Fix logical replication launcher wake up and reset After the logical replication launcher was told to wake up at commit (for example, by a CREATE SUBSCRIPTION command), the flag to wake up was not reset, so it would be woken up at every following commit as well. So fix that by resetting the flag. Also, we don't need to wake up anything if the transaction was rolled back. Just reset the flag in that case. Author: Masahiko Sawada <sawada.mshk@gmail.com> Reported-by: Fujii Masao <masao.fujii@gmail.com>
2017-05-01 16:18:09 +02:00
AtEOXact_ApplyLauncher(false);
Wake up a subscription's replication worker processes after DDL. Waken related worker processes immediately at commit of a transaction that has performed ALTER SUBSCRIPTION (including the RENAME and OWNER variants). This reduces the response time for such operations. In the real world that might not be worth much, but it shaves several seconds off the runtime for the subscription test suite. In the case of PREPARE, we just throw away this notification state; it doesn't seem worth the work to preserve it. The workers will still react after the eventual COMMIT PREPARED, but not as quickly. Nathan Bossart Discussion: https://postgr.es/m/20221122004119.GA132961@nathanxps13
2023-01-06 22:08:20 +01:00
AtEOXact_LogicalRepWorkers(false);
Fix assorted core dumps and Assert failures that could occur during AbortTransaction or AbortSubTransaction, when trying to clean up after an error that prevented (sub)transaction start from completing: * access to TopTransactionResourceOwner that might not exist * assert failure in AtEOXact_GUC, if AtStart_GUC not called yet * assert failure or core dump in AfterTriggerEndSubXact, if AfterTriggerBeginSubXact not called yet Per testing by injecting elog(ERROR) at successive steps in StartTransaction and StartSubTransaction. It's not clear whether all of these cases could really occur in the field, but at least one of them is easily exposed by simple stress testing, as per my accidental discovery yesterday.
2010-01-24 22:49:17 +01:00
pgstat_report_xact_timestamp(0);
}
Improve documentation about reasoning behind the order of operations in GetSnapshotData, GetNewTransactionId, CommitTransaction, AbortTransaction, etc. Correct race condition in transaction status testing in HeapTupleSatisfiesVacuum --- this wasn't important for old VACUUM with exclusive lock on its table, but it sure is important now. All per pghackers discussion 7/11/01 and 7/12/01.
2001-07-17 00:43:34 +02:00
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
* State remains TRANS_ABORT until CleanupTransaction().
*/
Make critical sections (elog->crash) and interrupt holdoff sections into distinct concepts, per recent discussion on pghackers.
2001-01-19 23:08:47 +01:00
RESUME_INTERRUPTS();
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* CleanupTransaction
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
*/
static void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
CleanupTransaction(void)
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
{
TransactionState s = CurrentTransactionState;
/*
* State should still be TRANS_ABORT from AbortTransaction().
*/
if (s->state != TRANS_ABORT)
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
elog(FATAL, "CleanupTransaction: unexpected state %s",
TransStateAsString(s->state));
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
* do abort cleanup processing
*/
Portal and memory management infrastructure for extended query protocol. Both plannable queries and utility commands are now always executed within Portals, which have been revamped so that they can handle the load (they used to be good only for single SELECT queries). Restructure code to push command-completion-tag selection logic out of postgres.c, so that it won't have to be duplicated between simple and extended queries. initdb forced due to addition of a field to Query nodes.
2003-05-02 22:54:36 +02:00
AtCleanup_Portals(); /* now safe to release portal memory */
Always SnapshotResetXmin() during ClearTransaction() Avoid corner cases during 2PC with 6bad580d9e678a0b604883e14d8401d469b06566
2017-04-06 16:30:22 +02:00
AtEOXact_Snapshot(false, true); /* and release the transaction's snapshots */
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
CurrentResourceOwner = NULL; /* and resource owner */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (TopTransactionResourceOwner)
ResourceOwnerDelete(TopTransactionResourceOwner);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
s->curTransactionOwner = NULL;
CurTransactionResourceOwner = NULL;
TopTransactionResourceOwner = NULL;
Portal and memory management infrastructure for extended query protocol. Both plannable queries and utility commands are now always executed within Portals, which have been revamped so that they can handle the load (they used to be good only for single SELECT queries). Restructure code to push command-completion-tag selection logic out of postgres.c, so that it won't have to be duplicated between simple and extended queries. initdb forced due to addition of a field to Query nodes.
2003-05-02 22:54:36 +02:00
AtCleanup_Memory(); /* and transaction memory */
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = InvalidFullTransactionId;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->subTransactionId = InvalidSubTransactionId;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->nestingLevel = 0;
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
s->gucNestLevel = 0;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
s->childXids = NULL;
s->nChildXids = 0;
s->maxChildXids = 0;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
s->parallelModeLevel = 0;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
XactTopFullTransactionId = InvalidFullTransactionId;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nParallelCurrentXids = 0;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* done with abort processing, set current transaction state back to
* default
*/
s->state = TRANS_DEFAULT;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* StartTransactionCommand
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
void
Backend support for autocommit removed, per recent discussions. The only remnant of this failed experiment is that the server will take SET AUTOCOMMIT TO ON. Still TODO: provide some client-side autocommit logic in libpq.
2003-05-14 05:26:03 +02:00
StartTransactionCommand(void)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
{
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
switch (s->blockState)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
{
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* if we aren't in a transaction block, we just do our usual start
* transaction.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_DEFAULT:
StartTransaction();
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
s->blockState = TBLOCK_STARTED;
break;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are somewhere in a transaction block or subtransaction and
* about to start a new command. For now we do nothing, but
* someday we may do command-local resource initialization. (Note
* that any needed CommandCounterIncrement was done by the
* previous CommitTransactionCommand.)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_INPROGRESS:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Here we are in a failed transaction block (one of the commands
* caused an abort) so we do nothing but remain in the abort
* state. Eventually we will get a ROLLBACK command which will
* get us out of this state. (It is up to other code to ensure
* that no commands other than ROLLBACK will be processed in these
* states.)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_ABORT:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* These cases are invalid. */
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBBEGIN:
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
elog(ERROR, "StartTransactionCommand: unexpected state %s",
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
BlockStateAsString(s->blockState));
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
/*
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
* We must switch to CurTransactionContext before returning. This is
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
* already done if we called StartTransaction, otherwise not.
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Assert(CurTransactionContext != NULL);
MemoryContextSwitchTo(CurTransactionContext);
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
/*
* Simple system for saving and restoring transaction characteristics
* (isolation level, read only, deferrable). We need this for transaction
* chaining, so that we can set the characteristics of the new transaction to
* be the same as the previous one. (We need something like this because the
* GUC system resets the characteristics at transaction end, so for example
* just skipping the reset in StartTransaction() won't work.)
*/
void
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
SaveTransactionCharacteristics(SavedTransactionCharacteristics *s)
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
{
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
s->save_XactIsoLevel = XactIsoLevel;
s->save_XactReadOnly = XactReadOnly;
s->save_XactDeferrable = XactDeferrable;
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
}
void
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
RestoreTransactionCharacteristics(const SavedTransactionCharacteristics *s)
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
{
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
XactIsoLevel = s->save_XactIsoLevel;
XactReadOnly = s->save_XactReadOnly;
XactDeferrable = s->save_XactDeferrable;
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* CommitTransactionCommand
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
void
Backend support for autocommit removed, per recent discussions. The only remnant of this failed experiment is that the server will take SET AUTOCOMMIT TO ON. Still TODO: provide some client-side autocommit logic in libpq.
2003-05-14 05:26:03 +02:00
CommitTransactionCommand(void)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
{
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
TransactionState s = CurrentTransactionState;
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
SavedTransactionCharacteristics savetc;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
SaveTransactionCharacteristics(&savetc);
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
switch (s->blockState)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
{
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
/*
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* These shouldn't happen. TBLOCK_DEFAULT means the previous
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
* StartTransactionCommand didn't set the STARTED state
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* appropriately, while TBLOCK_PARALLEL_INPROGRESS should be ended
Fix typos in comments. Backpatch to all supported versions, where applicable, to make backpatching of future fixes go more smoothly. Josh Soref Discussion: https://www.postgresql.org/message-id/CACZqfqCf+5qRztLPgmmosr-B0Ye4srWzzw_mo4c_8_B_mtjmJQ@mail.gmail.com
2017-02-06 10:33:58 +01:00
* by EndParallelWorkerTransaction(), not this function.
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
*/
case TBLOCK_DEFAULT:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
elog(FATAL, "CommitTransactionCommand: unexpected state %s",
BlockStateAsString(s->blockState));
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Backend support for autocommit removed, per recent discussions. The only remnant of this failed experiment is that the server will take SET AUTOCOMMIT TO ON. Still TODO: provide some client-side autocommit logic in libpq.
2003-05-14 05:26:03 +02:00
* If we aren't in a transaction block, just do our usual
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* transaction commit, and return to the idle state.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_STARTED:
Backend support for autocommit removed, per recent discussions. The only remnant of this failed experiment is that the server will take SET AUTOCOMMIT TO ON. Still TODO: provide some client-side autocommit logic in libpq.
2003-05-14 05:26:03 +02:00
CommitTransaction();
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
s->blockState = TBLOCK_DEFAULT;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are completing a "BEGIN TRANSACTION" command, so we change
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
* to the "transaction block in progress" state and return. (We
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* assume the BEGIN did nothing to the database, so we need no
* CommandCounterIncrement.)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_BEGIN:
s->blockState = TBLOCK_INPROGRESS;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* This is the case when we have finished executing a command
* someplace within a transaction block. We increment the command
Portal and memory management infrastructure for extended query protocol. Both plannable queries and utility commands are now always executed within Portals, which have been revamped so that they can handle the load (they used to be good only for single SELECT queries). Restructure code to push command-completion-tag selection logic out of postgres.c, so that it won't have to be duplicated between simple and extended queries. initdb forced due to addition of a field to Query nodes.
2003-05-02 22:54:36 +02:00
* counter and return.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_INPROGRESS:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBINPROGRESS:
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
CommandCounterIncrement();
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are completing a "COMMIT" command. Do it and return to the
* idle state.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_END:
CommitTransaction();
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
s->blockState = TBLOCK_DEFAULT;
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
if (s->chain)
{
StartTransaction();
s->blockState = TBLOCK_INPROGRESS;
s->chain = false;
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
RestoreTransactionCharacteristics(&savetc);
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
}
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* Here we are in the middle of a transaction block but one of the
* commands caused an abort so we do nothing but remain in the
Fix typos in comments Dmitriy Olshevskiy
2015-05-17 14:57:30 +02:00
* abort state. Eventually we will get a ROLLBACK command.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_ABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT:
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Here we were in an aborted transaction block and we just got
* the ROLLBACK command from the user, so clean up the
* already-aborted transaction and return to the idle state.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
CleanupTransaction();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
s->blockState = TBLOCK_DEFAULT;
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
if (s->chain)
{
StartTransaction();
s->blockState = TBLOCK_INPROGRESS;
s->chain = false;
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
RestoreTransactionCharacteristics(&savetc);
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
}
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Here we were in a perfectly good transaction block but the user
* told us to ROLLBACK anyway. We have to abort the transaction
* and then clean up.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_PENDING:
AbortTransaction();
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
if (s->chain)
{
StartTransaction();
s->blockState = TBLOCK_INPROGRESS;
s->chain = false;
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
RestoreTransactionCharacteristics(&savetc);
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
}
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
break;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* We are completing a "PREPARE TRANSACTION" command. Do it and
* return to the idle state.
*/
case TBLOCK_PREPARE:
PrepareTransaction();
s->blockState = TBLOCK_DEFAULT;
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
Pre-beta mechanical code beautification. Run pgindent, pgperltidy, and reformat-dat-files. This set of diffs is a bit larger than typical. We've updated to pg_bsd_indent 2.1.2, which properly indents variable declarations that have multi-line initialization expressions (the continuation lines are now indented one tab stop). We've also updated to perltidy version 20230309 and changed some of its settings, which reduces its desire to add whitespace to lines to make assignments etc. line up. Going forward, that should make for fewer random-seeming changes to existing code. Discussion: https://postgr.es/m/20230428092545.qfb3y5wcu4cm75ur@alvherre.pgsql
2023-05-19 23:24:48 +02:00
* The user issued a SAVEPOINT inside a transaction block. Start a
* subtransaction. (DefineSavepoint already did PushTransaction,
* so as to have someplace to put the SUBBEGIN state.)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
case TBLOCK_SUBBEGIN:
StartSubTransaction();
s->blockState = TBLOCK_SUBINPROGRESS;
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Rephrase comments to make them clearer The use of "we" when referring to the active backend might be misunderstood, so rephrase to make it clearer who is performing the actions discussed in the comment. Author: Junwang Zhao <zhjwpku@gmail.com> Reviewed-by: Erikjan Rijkers <er@xs4all.nl> Reviewed-by: Robert Treat <rob@xzilla.net> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Discussion: https://postgr.es/m/CAEG8a3LRSMqkvjiURiJoSi4aGWORpiXUmUfQQK5PaD6WfPzu3w@mail.gmail.com
2022-08-04 16:30:06 +02:00
* The user issued a RELEASE command, so we end the current
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
* subtransaction and return to the parent transaction. The parent
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
* might be ended too, so repeat till we find an INPROGRESS
* transaction or subtransaction.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
do
{
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
CommitSubTransaction();
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
s = CurrentTransactionState; /* changed by pop */
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
} while (s->blockState == TBLOCK_SUBRELEASE);
Assert(s->blockState == TBLOCK_INPROGRESS ||
s->blockState == TBLOCK_SUBINPROGRESS);
break;
/*
Rephrase comments to make them clearer The use of "we" when referring to the active backend might be misunderstood, so rephrase to make it clearer who is performing the actions discussed in the comment. Author: Junwang Zhao <zhjwpku@gmail.com> Reviewed-by: Erikjan Rijkers <er@xs4all.nl> Reviewed-by: Robert Treat <rob@xzilla.net> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Discussion: https://postgr.es/m/CAEG8a3LRSMqkvjiURiJoSi4aGWORpiXUmUfQQK5PaD6WfPzu3w@mail.gmail.com
2022-08-04 16:30:06 +02:00
* The user issued a COMMIT, so we end the current subtransaction
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
* hierarchy and perform final commit. We do this by rolling up
* any subtransactions into their parent, which leads to O(N^2)
* operations with respect to resource owners - this isn't that
* bad until we approach a thousands of savepoints but is
* necessary for correctness should after triggers create new
* resource owners.
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
*/
case TBLOCK_SUBCOMMIT:
do
{
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
CommitSubTransaction();
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
s = CurrentTransactionState; /* changed by pop */
} while (s->blockState == TBLOCK_SUBCOMMIT);
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
/* If we had a COMMIT command, finish off the main xact too */
if (s->blockState == TBLOCK_END)
{
Assert(s->parent == NULL);
CommitTransaction();
s->blockState = TBLOCK_DEFAULT;
Fix bug in COMMIT AND CHAIN command. This commit fixes COMMIT AND CHAIN command so that it starts new transaction immediately even if savepoints are defined within the transaction to commit. Previously COMMIT AND CHAIN command did not in that case because commit 280a408b48 forgot to make CommitTransactionCommand() handle a transaction chaining when the transaction state was TBLOCK_SUBCOMMIT. Also this commit adds the regression test for COMMIT AND CHAIN command when savepoints are defined. Back-patch to v12 where transaction chaining was added. Reported-by: Arthur Nascimento Author: Fujii Masao Reviewed-by: Arthur Nascimento, Vik Fearing Discussion: https://postgr.es/m/16867-3475744069228158@postgresql.org
2021-02-19 13:57:52 +01:00
if (s->chain)
{
StartTransaction();
s->blockState = TBLOCK_INPROGRESS;
s->chain = false;
Don't use static storage for SaveTransactionCharacteristics(). This is pretty queasy-making on general principles, and the more so once you notice that CommitTransactionCommand() is actually stomping on the values saved by _SPI_commit(). It's okay as long as the active values didn't change during HoldPinnedPortals(); but that's a larger assumption than I think we want to make, especially since the fix is so simple. Discussion: https://postgr.es/m/1533956.1645731245@sss.pgh.pa.us
2022-02-28 18:54:12 +01:00
RestoreTransactionCharacteristics(&savetc);
Fix bug in COMMIT AND CHAIN command. This commit fixes COMMIT AND CHAIN command so that it starts new transaction immediately even if savepoints are defined within the transaction to commit. Previously COMMIT AND CHAIN command did not in that case because commit 280a408b48 forgot to make CommitTransactionCommand() handle a transaction chaining when the transaction state was TBLOCK_SUBCOMMIT. Also this commit adds the regression test for COMMIT AND CHAIN command when savepoints are defined. Back-patch to v12 where transaction chaining was added. Reported-by: Arthur Nascimento Author: Fujii Masao Reviewed-by: Arthur Nascimento, Vik Fearing Discussion: https://postgr.es/m/16867-3475744069228158@postgresql.org
2021-02-19 13:57:52 +01:00
}
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
else if (s->blockState == TBLOCK_PREPARE)
{
Assert(s->parent == NULL);
PrepareTransaction();
s->blockState = TBLOCK_DEFAULT;
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
else
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
elog(ERROR, "CommitTransactionCommand: unexpected state %s",
BlockStateAsString(s->blockState));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* The current already-failed subtransaction is ending due to a
* ROLLBACK or ROLLBACK TO command, so pop it and recursively
* examine the parent (which could be in any of several states).
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_END:
CleanupSubTransaction();
CommitTransactionCommand();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* As above, but it's not dead yet, so abort first.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_PENDING:
AbortSubTransaction();
CleanupSubTransaction();
CommitTransactionCommand();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* The current subtransaction is the target of a ROLLBACK TO
* command. Abort and pop it, then start a new subtransaction
* with the same name.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
char *name;
int savepointLevel;
/* save name and keep Cleanup from freeing it */
name = s->name;
s->name = NULL;
savepointLevel = s->savepointLevel;
AbortSubTransaction();
CleanupSubTransaction();
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
DefineSavepoint(NULL);
s = CurrentTransactionState; /* changed by push */
s->name = name;
s->savepointLevel = savepointLevel;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/* This is the same as TBLOCK_SUBBEGIN case */
Remove AssertArg and AssertState These don't offer anything over plain Assert, and their usage had already been declared obsolescent. Author: Nathan Bossart <nathandbossart@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Discussion: https://www.postgresql.org/message-id/20221009210148.GA900071@nathanxps13
2022-10-28 09:19:06 +02:00
Assert(s->blockState == TBLOCK_SUBBEGIN);
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
StartSubTransaction();
s->blockState = TBLOCK_SUBINPROGRESS;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* Same as above, but the subtransaction had already failed, so we
* don't need AbortSubTransaction.
*/
case TBLOCK_SUBABORT_RESTART:
{
char *name;
int savepointLevel;
Pgindent run for 8.0.
2004-08-29 07:07:03 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* save name and keep Cleanup from freeing it */
name = s->name;
s->name = NULL;
savepointLevel = s->savepointLevel;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CleanupSubTransaction();
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
DefineSavepoint(NULL);
s = CurrentTransactionState; /* changed by push */
s->name = name;
s->savepointLevel = savepointLevel;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* This is the same as TBLOCK_SUBBEGIN case */
Remove AssertArg and AssertState These don't offer anything over plain Assert, and their usage had already been declared obsolescent. Author: Nathan Bossart <nathandbossart@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Discussion: https://www.postgresql.org/message-id/20221009210148.GA900071@nathanxps13
2022-10-28 09:19:06 +02:00
Assert(s->blockState == TBLOCK_SUBBEGIN);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
StartSubTransaction();
s->blockState = TBLOCK_SUBINPROGRESS;
}
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
}
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* AbortCurrentTransaction
*/
void
AbortCurrentTransaction(void)
{
TransactionState s = CurrentTransactionState;
switch (s->blockState)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_DEFAULT:
Add code to prevent transaction ID wraparound by enforcing a safe limit in GetNewTransactionId(). Since the limit value has to be computed before we run any real transactions, this requires adding code to database startup to scan pg_database and determine the oldest datfrozenxid. This can conveniently be combined with the first stage of an attack on the problem that the 'flat file' copies of pg_shadow and pg_group are not properly updated during WAL recovery. The code I've added to startup resides in a new file src/backend/utils/init/flatfiles.c, and it is responsible for rewriting the flat files as well as initializing the XID wraparound limit value. This will eventually allow us to get rid of GetRawDatabaseInfo too, but we'll need an initdb so we can add a trigger to pg_database.
2005-02-20 03:22:07 +01:00
if (s->state == TRANS_DEFAULT)
{
/* we are idle, so nothing to do */
}
else
{
/*
* We can get here after an error during transaction start
* (state will be TRANS_START). Need to clean up the
* incompletely started transaction. First, adjust the
* low-level state to suppress warning message from
* AbortTransaction.
*/
if (s->state == TRANS_START)
s->state = TRANS_INPROGRESS;
AbortTransaction();
CleanupTransaction();
}
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
* If we aren't in a transaction block, we just do the basic abort
* & cleanup transaction. For this purpose, we treat an implicit
* transaction block as if it were a simple statement.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_STARTED:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
AbortTransaction();
Backend support for autocommit removed, per recent discussions. The only remnant of this failed experiment is that the server will take SET AUTOCOMMIT TO ON. Still TODO: provide some client-side autocommit logic in libpq.
2003-05-14 05:26:03 +02:00
CleanupTransaction();
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
s->blockState = TBLOCK_DEFAULT;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
* If we are in TBLOCK_BEGIN it means something screwed up right
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* after reading "BEGIN TRANSACTION". We assume that the user
* will interpret the error as meaning the BEGIN failed to get him
* into a transaction block, so we should abort and return to idle
* state.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_BEGIN:
AbortTransaction();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are somewhere in a transaction block and we've gotten a
* failure, so we abort the transaction and set up the persistent
* ABORT state. We will stay in ABORT until we get a ROLLBACK.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_INPROGRESS:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
AbortTransaction();
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
s->blockState = TBLOCK_ABORT;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* CleanupTransaction happens when we exit TBLOCK_ABORT_END */
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Here, we failed while trying to COMMIT. Clean up the
* transaction and return to idle state (we do not want to stay in
* the transaction).
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_END:
AbortTransaction();
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
CleanupTransaction();
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
s->blockState = TBLOCK_DEFAULT;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
* Here, we are already in an aborted transaction state and are
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* waiting for a ROLLBACK, but for some reason we failed again! So
* we just remain in the abort state.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
case TBLOCK_ABORT:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
break;
Another PGINDENT run that changes variable indenting and case label indenting. Also static variable indenting.
1997-09-08 04:41:22 +02:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are in a failed transaction and we got the ROLLBACK command.
* We have already aborted, we just need to cleanup and go to idle
* state.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
First phase of memory management rewrite (see backend/utils/mmgr/README for details). It doesn't really do that much yet, since there are no short-term memory contexts in the executor, but the infrastructure is in place and long-term contexts are handled reasonably. A few long- standing bugs have been fixed, such as 'VACUUM; anything' in a single query string crashing. Also, out-of-memory is now considered a recoverable ERROR, not FATAL. Eliminate a large amount of crufty, now-dead code in and around memory management. Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and backend startup.
2000-06-28 05:33:33 +02:00
CleanupTransaction();
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
s->blockState = TBLOCK_DEFAULT;
break;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are in a live transaction and we got a ROLLBACK command.
* Abort, cleanup, go to idle state.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_PENDING:
AbortTransaction();
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* Here, we failed while trying to PREPARE. Clean up the
* transaction and return to idle state (we do not want to stay in
* the transaction).
*/
case TBLOCK_PREPARE:
AbortTransaction();
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* We got an error inside a subtransaction. Abort just the
* subtransaction, and go to the persistent SUBABORT state until
* we get ROLLBACK.
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
AbortSubTransaction();
s->blockState = TBLOCK_SUBABORT;
break;
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* If we failed while trying to create a subtransaction, clean up
* the broken subtransaction and abort the parent. The same
* applies if we get a failure while ending a subtransaction.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
AbortSubTransaction();
CleanupSubTransaction();
AbortCurrentTransaction();
break;
/*
* Same as above, except the Abort() was already done.
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_END:
case TBLOCK_SUBABORT_RESTART:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CleanupSubTransaction();
AbortCurrentTransaction();
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
* PreventInTransactionBlock
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* This routine is to be called by statements that must not run inside
* a transaction block, typically because they have non-rollback-able
* side effects or do internal commits.
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
*
Doc: add comments about PreventInTransactionBlock/IsInTransactionBlock. Add a little to the header comments for these functions to make it clearer what guarantees about commit behavior are provided to callers. (See commit f92944137 for context.) Although this is only a comment change, it's really documentation aimed at authors of extensions, so it seems appropriate to back-patch. Yugo Nagata and Tom Lane, per further discussion of bug #17434. Discussion: https://postgr.es/m/17434-d9f7a064ce2a88a3@postgresql.org
2022-11-09 17:08:52 +01:00
* If this routine completes successfully, then the calling statement is
* guaranteed that if it completes without error, its results will be
* committed immediately.
*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* If we have already started a transaction block, issue an error; also issue
* an error if we appear to be running inside a user-defined function (which
* could issue more commands and possibly cause a failure after the statement
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
* completes). Subtransactions are verboten too.
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
*
Force immediate commit after CREATE DATABASE etc in extended protocol. We have a few commands that "can't run in a transaction block", meaning that if they complete their processing but then we fail to COMMIT, we'll be left with inconsistent on-disk state. However, the existing defenses for this are only watertight for simple query protocol. In extended protocol, we didn't commit until receiving a Sync message. Since the client is allowed to issue another command instead of Sync, we're in trouble if that command fails or is an explicit ROLLBACK. In any case, sitting in an inconsistent state while waiting for a client message that might not come seems pretty risky. This case wasn't reachable via libpq before we introduced pipeline mode, but it's always been an intended aspect of extended query protocol, and likely there are other clients that could reach it before. To fix, set a flag in PreventInTransactionBlock that tells exec_execute_message to force an immediate commit. This seems to be the approach that does least damage to existing working cases while still preventing the undesirable outcomes. While here, add some documentation to protocol.sgml that explicitly says how to use pipelining. That's latent in the existing docs if you know what to look for, but it's better to spell it out; and it provides a place to document this new behavior. Per bug #17434 from Yugo Nagata. It's been wrong for ages, so back-patch to all supported branches. Discussion: https://postgr.es/m/17434-d9f7a064ce2a88a3@postgresql.org
2022-07-26 19:07:03 +02:00
* We must also set XACT_FLAGS_NEEDIMMEDIATECOMMIT in MyXactFlags, to ensure
* that postgres.c follows through by committing after the statement is done.
*
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
* isTopLevel: passed down from ProcessUtility to determine whether we are
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
* inside a function. (We will always fail if this is false, but it's
* convenient to centralize the check here instead of making callers do it.)
Arrange for PreventTransactionChain to reject commands submitted as part of a multi-statement simple-Query message. This bug goes all the way back, but unfortunately is not nearly so easy to fix in existing releases; it is only the recent ProcessUtility API change that makes it fixable in HEAD. Per report from William Garrison.
2007-03-22 20:55:04 +01:00
* stmtType: statement type name, for error messages.
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
*/
void
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
PreventInTransactionBlock(bool isTopLevel, const char *stmtType)
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
{
/*
* xact block already started?
*/
if (IsTransactionBlock())
Error message editing in backend/access.
2003-07-21 22:29:40 +02:00
ereport(ERROR,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s cannot run inside a transaction block",
stmtType)));
pgindent run.
2003-08-04 02:43:34 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* subtransaction?
*/
if (IsSubTransaction())
ereport(ERROR,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s cannot run inside a subtransaction",
stmtType)));
Rethink handling of [Prevent|Is]InTransactionBlock in pipeline mode. Commits f92944137 et al. made IsInTransactionBlock() set the XACT_FLAGS_NEEDIMMEDIATECOMMIT flag before returning "false", on the grounds that that kept its API promises equivalent to those of PreventInTransactionBlock(). This turns out to be a bad idea though, because it allows an ANALYZE in a pipelined series of commands to cause an immediate commit, which is unexpected. Furthermore, if we return "false" then we have another issue, which is that ANALYZE will decide it's allowed to do internal commit-and-start-transaction sequences, thus possibly unexpectedly committing the effects of previous commands in the pipeline. To fix the latter situation, invent another transaction state flag XACT_FLAGS_PIPELINING, which explicitly records the fact that we have executed some extended-protocol command and not yet seen a commit for it. Then, require that flag to not be set before allowing InTransactionBlock() to return "false". Having done that, we can remove its setting of NEEDIMMEDIATECOMMIT without fear of causing problems. This means that the API guarantees of IsInTransactionBlock now diverge from PreventInTransactionBlock, which is mildly annoying, but it seems OK given the very limited usage of IsInTransactionBlock. (In any case, a caller preferring the old behavior could always set NEEDIMMEDIATECOMMIT for itself.) For consistency also require XACT_FLAGS_PIPELINING to not be set in PreventInTransactionBlock. This too is meant to prevent commands such as CREATE DATABASE from silently committing previous commands in a pipeline. Per report from Peter Eisentraut. As before, back-patch to all supported branches (which sadly no longer includes v10). Discussion: https://postgr.es/m/65a899dd-aebc-f667-1d0a-abb89ff3abf8@enterprisedb.com
2022-12-13 20:23:58 +01:00
/*
* inside a pipeline that has started an implicit transaction?
*/
if (MyXactFlags & XACT_FLAGS_PIPELINING)
ereport(ERROR,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s cannot be executed within a pipeline",
stmtType)));
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
/*
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
* inside a function call?
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
*/
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
if (!isTopLevel)
Error message editing in backend/access.
2003-07-21 22:29:40 +02:00
ereport(ERROR,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
errmsg("%s cannot be executed from a function", stmtType)));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
/* If we got past IsTransactionBlock test, should be in default state */
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
if (CurrentTransactionState->blockState != TBLOCK_DEFAULT &&
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
CurrentTransactionState->blockState != TBLOCK_STARTED)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
elog(FATAL, "cannot prevent transaction chain");
Force immediate commit after CREATE DATABASE etc in extended protocol. We have a few commands that "can't run in a transaction block", meaning that if they complete their processing but then we fail to COMMIT, we'll be left with inconsistent on-disk state. However, the existing defenses for this are only watertight for simple query protocol. In extended protocol, we didn't commit until receiving a Sync message. Since the client is allowed to issue another command instead of Sync, we're in trouble if that command fails or is an explicit ROLLBACK. In any case, sitting in an inconsistent state while waiting for a client message that might not come seems pretty risky. This case wasn't reachable via libpq before we introduced pipeline mode, but it's always been an intended aspect of extended query protocol, and likely there are other clients that could reach it before. To fix, set a flag in PreventInTransactionBlock that tells exec_execute_message to force an immediate commit. This seems to be the approach that does least damage to existing working cases while still preventing the undesirable outcomes. While here, add some documentation to protocol.sgml that explicitly says how to use pipelining. That's latent in the existing docs if you know what to look for, but it's better to spell it out; and it provides a place to document this new behavior. Per bug #17434 from Yugo Nagata. It's been wrong for ages, so back-patch to all supported branches. Discussion: https://postgr.es/m/17434-d9f7a064ce2a88a3@postgresql.org
2022-07-26 19:07:03 +02:00
/* All okay. Set the flag to make sure the right thing happens later. */
MyXactFlags |= XACT_FLAGS_NEEDIMMEDIATECOMMIT;
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
}
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
/*
Fix more typos and inconsistencies in the tree Author: Alexander Lakhin Discussion: https://postgr.es/m/0a5419ea-1452-a4e6-72ff-545b1a5a8076@gmail.com
2019-06-17 09:13:16 +02:00
* WarnNoTransactionBlock
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
* RequireTransactionBlock
Update function comments After a6542a4b6870a019cd952d055d2e7af2da2fe102, some function comments were misplaced. Fix that. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:29:20 +01:00
*
* These two functions allow for warnings or errors if a command is executed
* outside of a transaction block. This is useful for commands that have no
* effects that persist past transaction end (and so calling them outside a
* transaction block is presumably an error). DECLARE CURSOR is an example.
* While top-level transaction control commands (BEGIN/COMMIT/ABORT) and SET
* that have no effect issue warnings, all other no-effect commands generate
* errors.
*
* If we appear to be running inside a user-defined function, we do not
* issue anything, since the function could issue more commands that make
* use of the current statement's results. Likewise subtransactions.
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
* Thus these are inverses for PreventInTransactionBlock.
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
*
Update function comments After a6542a4b6870a019cd952d055d2e7af2da2fe102, some function comments were misplaced. Fix that. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:29:20 +01:00
* isTopLevel: passed down from ProcessUtility to determine whether we are
* inside a function.
* stmtType: statement type name, for warning or error messages.
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
*/
void
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
WarnNoTransactionBlock(bool isTopLevel, const char *stmtType)
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
{
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
CheckTransactionBlock(isTopLevel, false, stmtType);
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
}
void
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
RequireTransactionBlock(bool isTopLevel, const char *stmtType)
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
{
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
CheckTransactionBlock(isTopLevel, true, stmtType);
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update function comments After a6542a4b6870a019cd952d055d2e7af2da2fe102, some function comments were misplaced. Fix that. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:29:20 +01:00
* This is the implementation of the above two.
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
*/
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
static void
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
CheckTransactionBlock(bool isTopLevel, bool throwError, const char *stmtType)
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
{
/*
* xact block already started?
*/
if (IsTransactionBlock())
return;
pgindent run.
2003-08-04 02:43:34 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* subtransaction?
*/
if (IsSubTransaction())
return;
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
/*
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
* inside a function call?
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
*/
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
if (!isTopLevel)
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
return;
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
Change SET LOCAL/CONSTRAINTS/TRANSACTION and ABORT behavior Change SET LOCAL/CONSTRAINTS/TRANSACTION behavior outside of a transaction block from error (post-9.3) to warning. (Was nothing in <= 9.3.) Also change ABORT outside of a transaction block from notice to warning.
2013-11-26 01:19:40 +01:00
ereport(throwError ? ERROR : WARNING,
Error message editing in backend/access.
2003-07-21 22:29:40 +02:00
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
Wording cleanup for error messages. Also change can't -> cannot. Standard English uses "may", "can", and "might" in different ways: may - permission, "You may borrow my rake." can - ability, "I can lift that log." might - possibility, "It might rain today." Unfortunately, in conversational English, their use is often mixed, as in, "You may use this variable to do X", when in fact, "can" is a better choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 20:10:30 +01:00
errmsg("%s can only be used in transaction blocks",
Error message editing in backend/access.
2003-07-21 22:29:40 +02:00
stmtType)));
Put back error test for DECLARE CURSOR outside a transaction block ... but do it correctly now.
2002-11-18 02:17:39 +01:00
}
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
/*
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
* IsInTransactionBlock
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
*
* This routine is for statements that need to behave differently inside
* a transaction block than when running as single commands. ANALYZE is
* currently the only example.
*
Rethink handling of [Prevent|Is]InTransactionBlock in pipeline mode. Commits f92944137 et al. made IsInTransactionBlock() set the XACT_FLAGS_NEEDIMMEDIATECOMMIT flag before returning "false", on the grounds that that kept its API promises equivalent to those of PreventInTransactionBlock(). This turns out to be a bad idea though, because it allows an ANALYZE in a pipelined series of commands to cause an immediate commit, which is unexpected. Furthermore, if we return "false" then we have another issue, which is that ANALYZE will decide it's allowed to do internal commit-and-start-transaction sequences, thus possibly unexpectedly committing the effects of previous commands in the pipeline. To fix the latter situation, invent another transaction state flag XACT_FLAGS_PIPELINING, which explicitly records the fact that we have executed some extended-protocol command and not yet seen a commit for it. Then, require that flag to not be set before allowing InTransactionBlock() to return "false". Having done that, we can remove its setting of NEEDIMMEDIATECOMMIT without fear of causing problems. This means that the API guarantees of IsInTransactionBlock now diverge from PreventInTransactionBlock, which is mildly annoying, but it seems OK given the very limited usage of IsInTransactionBlock. (In any case, a caller preferring the old behavior could always set NEEDIMMEDIATECOMMIT for itself.) For consistency also require XACT_FLAGS_PIPELINING to not be set in PreventInTransactionBlock. This too is meant to prevent commands such as CREATE DATABASE from silently committing previous commands in a pipeline. Per report from Peter Eisentraut. As before, back-patch to all supported branches (which sadly no longer includes v10). Discussion: https://postgr.es/m/65a899dd-aebc-f667-1d0a-abb89ff3abf8@enterprisedb.com
2022-12-13 20:23:58 +01:00
* If this routine returns "false", then the calling statement is allowed
* to perform internal transaction-commit-and-start cycles; there is not a
* risk of messing up any transaction already in progress. (Note that this
* is not the identical guarantee provided by PreventInTransactionBlock,
* since we will not force a post-statement commit.)
Doc: add comments about PreventInTransactionBlock/IsInTransactionBlock. Add a little to the header comments for these functions to make it clearer what guarantees about commit behavior are provided to callers. (See commit f92944137 for context.) Although this is only a comment change, it's really documentation aimed at authors of extensions, so it seems appropriate to back-patch. Yugo Nagata and Tom Lane, per further discussion of bug #17434. Discussion: https://postgr.es/m/17434-d9f7a064ce2a88a3@postgresql.org
2022-11-09 17:08:52 +01:00
*
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
* isTopLevel: passed down from ProcessUtility to determine whether we are
* inside a function.
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
*/
bool
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
IsInTransactionBlock(bool isTopLevel)
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
{
/*
Rename TransactionChain functions We call this thing a "transaction block" everywhere except in a few functions, where it is mysteriously called a "transaction chain". In the SQL standard, a transaction chain is something different. So rename these functions to match the common terminology. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:44:15 +01:00
* Return true on same conditions that would make
* PreventInTransactionBlock error out
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
*/
if (IsTransactionBlock())
return true;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
if (IsSubTransaction())
return true;
Rethink handling of [Prevent|Is]InTransactionBlock in pipeline mode. Commits f92944137 et al. made IsInTransactionBlock() set the XACT_FLAGS_NEEDIMMEDIATECOMMIT flag before returning "false", on the grounds that that kept its API promises equivalent to those of PreventInTransactionBlock(). This turns out to be a bad idea though, because it allows an ANALYZE in a pipelined series of commands to cause an immediate commit, which is unexpected. Furthermore, if we return "false" then we have another issue, which is that ANALYZE will decide it's allowed to do internal commit-and-start-transaction sequences, thus possibly unexpectedly committing the effects of previous commands in the pipeline. To fix the latter situation, invent another transaction state flag XACT_FLAGS_PIPELINING, which explicitly records the fact that we have executed some extended-protocol command and not yet seen a commit for it. Then, require that flag to not be set before allowing InTransactionBlock() to return "false". Having done that, we can remove its setting of NEEDIMMEDIATECOMMIT without fear of causing problems. This means that the API guarantees of IsInTransactionBlock now diverge from PreventInTransactionBlock, which is mildly annoying, but it seems OK given the very limited usage of IsInTransactionBlock. (In any case, a caller preferring the old behavior could always set NEEDIMMEDIATECOMMIT for itself.) For consistency also require XACT_FLAGS_PIPELINING to not be set in PreventInTransactionBlock. This too is meant to prevent commands such as CREATE DATABASE from silently committing previous commands in a pipeline. Per report from Peter Eisentraut. As before, back-patch to all supported branches (which sadly no longer includes v10). Discussion: https://postgr.es/m/65a899dd-aebc-f667-1d0a-abb89ff3abf8@enterprisedb.com
2022-12-13 20:23:58 +01:00
if (MyXactFlags & XACT_FLAGS_PIPELINING)
return true;
First phase of plan-invalidation project: create a plan cache management module and teach PREPARE and protocol-level prepared statements to use it. In service of this, rearrange utility-statement processing so that parse analysis does not assume table schemas can't change before execution for utility statements (necessary because we don't attempt to re-acquire locks for utility statements when reusing a stored plan). This requires some refactoring of the ProcessUtility API, but it ends up cleaner anyway, for instance we can get rid of the QueryContext global. Still to do: fix up SPI and related code to use the plan cache; I'm tempted to try to make SQL functions use it too. Also, there are at least some aspects of system state that we want to ensure remain the same during a replan as in the original processing; search_path certainly ought to behave that way for instance, and perhaps there are others.
2007-03-13 01:33:44 +01:00
if (!isTopLevel)
For multi-table ANALYZE, use per-table transactions when possible (ie, when not inside a transaction block), so that we can avoid holding locks longer than necessary. Per trouble report from Philip Warner.
2004-05-23 01:14:38 +02:00
return true;
if (CurrentTransactionState->blockState != TBLOCK_DEFAULT &&
CurrentTransactionState->blockState != TBLOCK_STARTED)
return true;
return false;
}
Fix places that were using IsTransactionBlock() as an (inadequate) check that they'd get to commit immediately on finishing. There's now a centralized routine PreventTransactionChain() that implements the necessary tests.
2002-10-22 00:06:20 +02:00
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
/*
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
* Register or deregister callback functions for start- and end-of-xact
* operations.
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
*
* These functions are intended for use by dynamically loaded modules.
* For built-in modules we generally just hardwire the appropriate calls
* (mainly because it's easier to control the order that way, where needed).
*
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
* At transaction end, the callback occurs post-commit or post-abort, so the
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* callback functions can only do noncritical cleanup.
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
*/
void
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
RegisterXactCallback(XactCallback callback, void *arg)
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
{
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
XactCallbackItem *item;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
item = (XactCallbackItem *)
MemoryContextAlloc(TopMemoryContext, sizeof(XactCallbackItem));
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
item->callback = callback;
item->arg = arg;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
item->next = Xact_callbacks;
Xact_callbacks = item;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
}
void
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
UnregisterXactCallback(XactCallback callback, void *arg)
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
{
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
XactCallbackItem *item;
XactCallbackItem *prev;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
prev = NULL;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
for (item = Xact_callbacks; item; prev = item, item = item->next)
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
{
if (item->callback == callback && item->arg == arg)
{
if (prev)
prev->next = item->next;
else
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
Xact_callbacks = item->next;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
pfree(item);
break;
}
}
}
static void
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CallXactCallbacks(XactEvent event)
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
{
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
XactCallbackItem *item;
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
XactCallbackItem *next;
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
for (item = Xact_callbacks; item; item = next)
{
/* allow callbacks to unregister themselves when called */
next = item->next;
Reduce excessive dereferencing of function pointers It is equivalent in ANSI C to write (*funcptr) () and funcptr(). These two styles have been applied inconsistently. After discussion, we'll use the more verbose style for plain function pointer variables, to make it clear that it's a variable, and the shorter style when the function pointer is in a struct (s.func() or s->func()), because then it's clear that it's not a plain function name, and otherwise the excessive punctuation makes some of those invocations hard to read. Discussion: https://www.postgresql.org/message-id/f52c16db-14ed-757d-4b48-7ef360b1631d@2ndquadrant.com
2017-09-07 18:06:23 +02:00
item->callback(event, item->arg);
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
}
/*
* Register or deregister callback functions for start- and end-of-subxact
* operations.
*
* Pretty much same as above, but for subtransaction events.
*
* At subtransaction end, the callback occurs post-subcommit or post-subabort,
* so the callback functions can only do noncritical cleanup. At
* subtransaction start, the callback is called when the subtransaction has
* finished initializing.
*/
void
RegisterSubXactCallback(SubXactCallback callback, void *arg)
{
SubXactCallbackItem *item;
item = (SubXactCallbackItem *)
MemoryContextAlloc(TopMemoryContext, sizeof(SubXactCallbackItem));
item->callback = callback;
item->arg = arg;
item->next = SubXact_callbacks;
SubXact_callbacks = item;
}
void
UnregisterSubXactCallback(SubXactCallback callback, void *arg)
{
SubXactCallbackItem *item;
SubXactCallbackItem *prev;
prev = NULL;
for (item = SubXact_callbacks; item; prev = item, item = item->next)
{
if (item->callback == callback && item->arg == arg)
{
if (prev)
prev->next = item->next;
else
SubXact_callbacks = item->next;
pfree(item);
break;
}
}
}
static void
CallSubXactCallbacks(SubXactEvent event,
SubTransactionId mySubid,
SubTransactionId parentSubid)
{
SubXactCallbackItem *item;
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
SubXactCallbackItem *next;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
for (item = SubXact_callbacks; item; item = next)
{
/* allow callbacks to unregister themselves when called */
next = item->next;
Reduce excessive dereferencing of function pointers It is equivalent in ANSI C to write (*funcptr) () and funcptr(). These two styles have been applied inconsistently. After discussion, we'll use the more verbose style for plain function pointer variables, to make it clear that it's a variable, and the shorter style when the function pointer is in a struct (s.func() or s->func()), because then it's clear that it's not a plain function name, and otherwise the excessive punctuation makes some of those invocations hard to read. Discussion: https://www.postgresql.org/message-id/f52c16db-14ed-757d-4b48-7ef360b1631d@2ndquadrant.com
2017-09-07 18:06:23 +02:00
item->callback(event, mySubid, parentSubid, item->arg);
Allow callback functions to deregister themselves during a call. Fetch the next-item pointer before the call not after, so that we aren't dereferencing a dangling pointer if the callback deregistered itself during the call. The risky coding pattern appears in CallXactCallbacks, CallSubXactCallbacks, and ResourceOwnerReleaseInternal. (There are some other places that might be at hazard if they offered deregistration functionality, but they don't.) I (tgl) considered back-patching this, but desisted because it wouldn't be very safe for extensions to rely on this working in pre-v16 branches. Hao Wu Discussion: https://postgr.es/m/CAH+9SWXTiERkmhRke+QCcc+jRH8d5fFHTxh8ZK0-Yn4BSpyaAg@mail.gmail.com
2022-09-28 17:23:14 +02:00
}
Add a mechanism to let dynamically loaded modules register post-commit/ post-abort cleanup hooks. I'm surprised that we have not needed this already, but I need it now to fix a plpgsql problem, and the usefulness for other dynamically loaded modules seems obvious.
2003-09-29 01:26:20 +02:00
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/* ----------------------------------------------------------------
* transaction block support
* ----------------------------------------------------------------
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* BeginTransactionBlock
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
* This executes a BEGIN command.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
void
All external function definitions now have prototypes that are checked.
1996-11-10 04:06:38 +01:00
BeginTransactionBlock(void)
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
{
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
/*
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
* We are not inside a transaction block, so allow one to begin.
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
*/
case TBLOCK_STARTED:
s->blockState = TBLOCK_BEGIN;
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
/*
* BEGIN converts an implicit transaction block to a regular one.
* (Note that we allow this even if we've already done some
* commands, which is a bit odd but matches historical practice.)
*/
case TBLOCK_IMPLICIT_INPROGRESS:
s->blockState = TBLOCK_BEGIN;
break;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* Already a transaction block in progress.
*/
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_INPROGRESS:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_ABORT:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
ereport(WARNING,
(errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
errmsg("there is already a transaction in progress")));
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_BEGIN:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBBEGIN:
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
case TBLOCK_SUBABORT_PENDING:
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
elog(FATAL, "BeginTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/*
* PrepareTransactionBlock
* This executes a PREPARE command.
*
* Since PREPARE may actually do a ROLLBACK, the result indicates what
Change TRUE/FALSE to true/false The lower case spellings are C and C++ standard and are used in most parts of the PostgreSQL sources. The upper case spellings are only used in some files/modules. So standardize on the standard spellings. The APIs for ICU, Perl, and Windows define their own TRUE and FALSE, so those are left as is when using those APIs. In code comments, we use the lower-case spelling for the C concepts and keep the upper-case spelling for the SQL concepts. Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
2017-08-16 06:22:32 +02:00
* happened: true for PREPARE, false for ROLLBACK.
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
*
* Note that we don't actually do anything here except change blockState.
* The real work will be done in the upcoming PrepareTransaction().
* We do it this way because it's not convenient to change memory context,
* resource owner, etc while executing inside a Portal.
*/
bool
Add some const decorations to prototypes Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2017-10-31 15:34:31 +01:00
PrepareTransactionBlock(const char *gid)
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
{
TransactionState s;
bool result;
/* Set up to commit the current transaction */
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
result = EndTransactionBlock(false);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* If successful, change outer tblock state to PREPARE */
if (result)
{
s = CurrentTransactionState;
while (s->parent != NULL)
s = s->parent;
if (s->blockState == TBLOCK_END)
{
/* Save GID where PrepareTransaction can find it again */
prepareGID = MemoryContextStrdup(TopTransactionContext, gid);
s->blockState = TBLOCK_PREPARE;
}
else
{
/*
* ignore case where we are not in a transaction;
* EndTransactionBlock already issued a warning.
*/
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
Assert(s->blockState == TBLOCK_STARTED ||
s->blockState == TBLOCK_IMPLICIT_INPROGRESS);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* Don't send back a PREPARE result tag... */
result = false;
}
}
return result;
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* EndTransactionBlock
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
* This executes a COMMIT command.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*
* Since COMMIT may actually do a ROLLBACK, the result indicates what
Change TRUE/FALSE to true/false The lower case spellings are C and C++ standard and are used in most parts of the PostgreSQL sources. The upper case spellings are only used in some files/modules. So standardize on the standard spellings. The APIs for ICU, Perl, and Windows define their own TRUE and FALSE, so those are left as is when using those APIs. In code comments, we use the lower-case spelling for the C concepts and keep the upper-case spelling for the SQL concepts. Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
2017-08-16 06:22:32 +02:00
* happened: true for COMMIT, false for ROLLBACK.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* Note that we don't actually do anything here except change blockState.
* The real work will be done in the upcoming CommitTransactionCommand().
* We do it this way because it's not convenient to change memory context,
* resource owner, etc while executing inside a Portal.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
bool
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
EndTransactionBlock(bool chain)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
bool result = false;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are in a transaction block, so tell CommitTransactionCommand
* to COMMIT.
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
*/
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_INPROGRESS:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
s->blockState = TBLOCK_END;
result = true;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
/*
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
* We are in an implicit transaction block. If AND CHAIN was
* specified, error. Otherwise commit, but issue a warning
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
* because there was no explicit BEGIN before this.
*/
case TBLOCK_IMPLICIT_INPROGRESS:
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
if (chain)
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"COMMIT AND CHAIN")));
else
ereport(WARNING,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
errmsg("there is no transaction in progress")));
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
s->blockState = TBLOCK_END;
result = true;
break;
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are in a failed transaction block. Tell
* CommitTransactionCommand it's time to exit the block.
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
*/
case TBLOCK_ABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->blockState = TBLOCK_ABORT_END;
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
/*
* We are in a live subtransaction block. Set up to subcommit all
* open subtransactions and then commit the main transaction.
*/
case TBLOCK_SUBINPROGRESS:
while (s->parent != NULL)
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (s->blockState == TBLOCK_SUBINPROGRESS)
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
s->blockState = TBLOCK_SUBCOMMIT;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
else
elog(FATAL, "EndTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
s = s->parent;
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (s->blockState == TBLOCK_INPROGRESS)
s->blockState = TBLOCK_END;
else
elog(FATAL, "EndTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
result = true;
break;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Here we are inside an aborted subtransaction. Treat the COMMIT
* as ROLLBACK: set up to abort everything and exit the main
* transaction.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
case TBLOCK_SUBABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
while (s->parent != NULL)
{
if (s->blockState == TBLOCK_SUBINPROGRESS)
s->blockState = TBLOCK_SUBABORT_PENDING;
else if (s->blockState == TBLOCK_SUBABORT)
s->blockState = TBLOCK_SUBABORT_END;
else
elog(FATAL, "EndTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
s = s->parent;
}
if (s->blockState == TBLOCK_INPROGRESS)
s->blockState = TBLOCK_ABORT_PENDING;
else if (s->blockState == TBLOCK_ABORT)
s->blockState = TBLOCK_ABORT_END;
else
elog(FATAL, "EndTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
/*
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
* The user issued COMMIT when not inside a transaction. For
* COMMIT without CHAIN, issue a WARNING, staying in
* TBLOCK_STARTED state. The upcoming call to
Change COMMIT back to the old behavior of emitting command tag COMMIT, not ROLLBACK, for the case of COMMIT outside a transaction block. Alvaro Herrera
2004-10-30 22:44:43 +02:00
* CommitTransactionCommand() will then close the transaction and
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
* put us back into the default state. For COMMIT AND CHAIN,
* error.
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_STARTED:
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
if (chain)
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"COMMIT AND CHAIN")));
else
ereport(WARNING,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
errmsg("there is no transaction in progress")));
Change COMMIT back to the old behavior of emitting command tag COMMIT, not ROLLBACK, for the case of COMMIT outside a transaction block. Alvaro Herrera
2004-10-30 22:44:43 +02:00
result = true;
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* The user issued a COMMIT that somehow ran inside a parallel
* worker. We can't cope with that.
*/
case TBLOCK_PARALLEL_INPROGRESS:
ereport(FATAL,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot commit during a parallel operation")));
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_BEGIN:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBBEGIN:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
elog(FATAL, "EndTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
Assert(s->blockState == TBLOCK_STARTED ||
s->blockState == TBLOCK_END ||
s->blockState == TBLOCK_ABORT_END ||
s->blockState == TBLOCK_ABORT_PENDING);
s->chain = chain;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
return result;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* UserAbortTransactionBlock
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
* This executes a ROLLBACK command.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* As above, we don't actually do anything here except change blockState.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
void
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
UserAbortTransactionBlock(bool chain)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are inside a transaction block and we got a ROLLBACK command
* from the user, so tell CommitTransactionCommand to abort and
* exit the transaction block.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_INPROGRESS:
s->blockState = TBLOCK_ABORT_PENDING;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are inside a failed transaction block and we got a ROLLBACK
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
* command from the user. Abort processing is already done, so
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* CommitTransactionCommand just has to cleanup and go back to
* idle state.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT:
s->blockState = TBLOCK_ABORT_END;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We are inside a subtransaction. Mark everything up to top
* level as exitable.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT:
while (s->parent != NULL)
{
if (s->blockState == TBLOCK_SUBINPROGRESS)
s->blockState = TBLOCK_SUBABORT_PENDING;
else if (s->blockState == TBLOCK_SUBABORT)
s->blockState = TBLOCK_SUBABORT_END;
else
elog(FATAL, "UserAbortTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
s = s->parent;
}
if (s->blockState == TBLOCK_INPROGRESS)
s->blockState = TBLOCK_ABORT_PENDING;
else if (s->blockState == TBLOCK_ABORT)
s->blockState = TBLOCK_ABORT_END;
else
elog(FATAL, "UserAbortTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
/*
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
* The user issued ABORT when not inside a transaction. For
* ROLLBACK without CHAIN, issue a WARNING and go to abort state.
* The upcoming call to CommitTransactionCommand() will then put
* us back into the default state. For ROLLBACK AND CHAIN, error.
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
*
* We do the same thing with ABORT inside an implicit transaction,
* although in this case we might be rolling back actual database
* state changes. (It's debatable whether we should issue a
* WARNING in this case, but we have done so historically.)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
case TBLOCK_STARTED:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Fix behavior of AND CHAIN outside of explicit transaction blocks When using COMMIT AND CHAIN or ROLLBACK AND CHAIN not in an explicit transaction block, the previous implementation would leave a transaction block active in the ROLLBACK case but not the COMMIT case. To fix for now, error out when using these commands not in an explicit transaction block. This restriction could be lifted if a sensible definition and implementation is found. Bug: #15977 Author: fn ln <emuser20140816@gmail.com> Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2019-09-08 16:11:21 +02:00
if (chain)
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"ROLLBACK AND CHAIN")));
else
ereport(WARNING,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
errmsg("there is no transaction in progress")));
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->blockState = TBLOCK_ABORT_PENDING;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* The user issued an ABORT that somehow ran inside a parallel
* worker. We can't cope with that.
*/
case TBLOCK_PARALLEL_INPROGRESS:
ereport(FATAL,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot abort during a parallel operation")));
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/* These cases are invalid. */
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_BEGIN:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
elog(FATAL, "UserAbortTransactionBlock: unexpected state %s",
BlockStateAsString(s->blockState));
break;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
}
Transaction chaining Add command variants COMMIT AND CHAIN and ROLLBACK AND CHAIN, which start new transactions with the same transaction characteristics as the just finished one, per SQL standard. Support for transaction chaining in PL/pgSQL is also added. This functionality is especially useful when running COMMIT in a loop in PL/pgSQL. Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr> Discussion: https://www.postgresql.org/message-id/flat/28536681-324b-10dc-ade8-ab46f7645a5a@2ndquadrant.com
2019-03-24 10:33:14 +01:00
Assert(s->blockState == TBLOCK_ABORT_END ||
s->blockState == TBLOCK_ABORT_PENDING);
s->chain = chain;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
}
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
/*
* BeginImplicitTransactionBlock
* Start an implicit transaction block if we're not already in one.
*
* Unlike BeginTransactionBlock, this is called directly from the main loop
* in postgres.c, not within a Portal. So we can just change blockState
* without a lot of ceremony. We do not expect caller to do
* CommitTransactionCommand/StartTransactionCommand.
*/
void
BeginImplicitTransactionBlock(void)
{
TransactionState s = CurrentTransactionState;
/*
* If we are in STARTED state (that is, no transaction block is open),
* switch to IMPLICIT_INPROGRESS state, creating an implicit transaction
* block.
*
* For caller convenience, we consider all other transaction states as
* legal here; otherwise the caller would need its own state check, which
* seems rather pointless.
*/
if (s->blockState == TBLOCK_STARTED)
s->blockState = TBLOCK_IMPLICIT_INPROGRESS;
}
/*
* EndImplicitTransactionBlock
* End an implicit transaction block, if we're in one.
*
* Like EndTransactionBlock, we just make any needed blockState change here.
* The real work will be done in the upcoming CommitTransactionCommand().
*/
void
EndImplicitTransactionBlock(void)
{
TransactionState s = CurrentTransactionState;
/*
* If we are in IMPLICIT_INPROGRESS state, switch back to STARTED state,
* allowing CommitTransactionCommand to commit whatever happened during
* the implicit transaction block as though it were a single statement.
*
* For caller convenience, we consider all other transaction states as
* legal here; otherwise the caller would need its own state check, which
* seems rather pointless.
*/
if (s->blockState == TBLOCK_IMPLICIT_INPROGRESS)
s->blockState = TBLOCK_STARTED;
}
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
* DefineSavepoint
* This executes a SAVEPOINT command.
*/
void
Add some const decorations to prototypes Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2017-10-31 15:34:31 +01:00
DefineSavepoint(const char *name)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
TransactionState s = CurrentTransactionState;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for new subtransactions after that
* point. (Note that this check will certainly error out if s->blockState
* is TBLOCK_PARALLEL_INPROGRESS, so we can treat that as an invalid case
* below.)
*/
if (IsInParallelMode())
ereport(ERROR,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot define savepoints during a parallel operation")));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
case TBLOCK_INPROGRESS:
case TBLOCK_SUBINPROGRESS:
/* Normal subtransaction start */
PushTransaction();
s = CurrentTransactionState; /* changed by push */
Pgindent run for 8.0.
2004-08-29 07:07:03 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Savepoint names, like the TransactionState block itself, live
* in TopTransactionContext.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (name)
s->name = MemoryContextStrdup(TopTransactionContext, name);
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
break;
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
/*
* We disallow savepoint commands in implicit transaction blocks.
* There would be no great difficulty in allowing them so far as
* this module is concerned, but a savepoint seems inconsistent
* with exec_simple_query's behavior of abandoning the whole query
* string upon error. Also, the point of an implicit transaction
* block (as opposed to a regular one) is to automatically close
* after an error, so it's hard to see how a savepoint would fit
* into that.
*
* The error messages for this are phrased as if there were no
* active transaction block at all, which is historical but
* perhaps could be improved.
*/
case TBLOCK_IMPLICIT_INPROGRESS:
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"SAVEPOINT")));
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBBEGIN:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_ABORT:
case TBLOCK_SUBABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
elog(FATAL, "DefineSavepoint: unexpected state %s",
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
BlockStateAsString(s->blockState));
break;
}
}
/*
* ReleaseSavepoint
* This executes a RELEASE command.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* As above, we don't actually do anything here except change blockState.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
void
Simplify parse representation of savepoint commands Instead of embedding the savepoint name in a list and then requiring complex code to unpack it, just add another struct field to store it directly. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:57:06 +01:00
ReleaseSavepoint(const char *name)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
TransactionState s = CurrentTransactionState;
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
TransactionState target,
xact;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for transaction state change after that
* point. (Note that this check will certainly error out if s->blockState
* is TBLOCK_PARALLEL_INPROGRESS, so we can treat that as an invalid case
* below.)
*/
if (IsInParallelMode())
ereport(ERROR,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot release savepoints during a parallel operation")));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
* We can't release a savepoint if there is no savepoint defined.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*/
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_INPROGRESS:
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist", name)));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
break;
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
/* See comment about implicit transactions in DefineSavepoint */
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"RELEASE SAVEPOINT")));
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
* We are in a non-aborted subtransaction. This is the only valid
* case.
*/
case TBLOCK_SUBINPROGRESS:
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBBEGIN:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT:
case TBLOCK_SUBABORT:
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
elog(FATAL, "ReleaseSavepoint: unexpected state %s",
BlockStateAsString(s->blockState));
break;
}
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
for (target = s; PointerIsValid(target); target = target->parent)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
if (PointerIsValid(target->name) && strcmp(target->name, name) == 0)
break;
}
if (!PointerIsValid(target))
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist", name)));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
/* disallow crossing savepoint level boundaries */
if (target->savepointLevel != s->savepointLevel)
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist within current savepoint level", name)));
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
/*
* Mark "commit pending" all subtransactions up to the target
* subtransaction. The actual commits will happen when control gets to
* CommitTransactionCommand.
*/
xact = CurrentTransactionState;
for (;;)
{
Assert(xact->blockState == TBLOCK_SUBINPROGRESS);
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
xact->blockState = TBLOCK_SUBRELEASE;
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
if (xact == target)
break;
xact = xact->parent;
Assert(PointerIsValid(xact));
}
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
}
/*
* RollbackToSavepoint
* This executes a ROLLBACK TO <savepoint> command.
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* As above, we don't actually do anything here except change blockState.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
void
Simplify parse representation of savepoint commands Instead of embedding the savepoint name in a list and then requiring complex code to unpack it, just add another struct field to store it directly. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 02:57:06 +01:00
RollbackToSavepoint(const char *name)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
TransactionState s = CurrentTransactionState;
TransactionState target,
xact;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for transaction state change after that
* point. (Note that this check will certainly error out if s->blockState
* is TBLOCK_PARALLEL_INPROGRESS, so we can treat that as an invalid case
* below.)
*/
if (IsInParallelMode())
ereport(ERROR,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot rollback to savepoints during a parallel operation")));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (s->blockState)
{
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* We can't rollback to a savepoint if there is no savepoint
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
* defined.
*/
case TBLOCK_INPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist", name)));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
break;
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
/* See comment about implicit transactions in DefineSavepoint */
ereport(ERROR,
(errcode(ERRCODE_NO_ACTIVE_SQL_TRANSACTION),
/* translator: %s represents an SQL statement name */
errmsg("%s can only be used in transaction blocks",
"ROLLBACK TO SAVEPOINT")));
break;
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
* There is at least one savepoint, so proceed.
*/
case TBLOCK_SUBINPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
elog(FATAL, "RollbackToSavepoint: unexpected state %s",
BlockStateAsString(s->blockState));
break;
}
Fix misplacement of savepointLevel test, per report from Chris K-L.
2004-08-03 17:57:26 +02:00
for (target = s; PointerIsValid(target); target = target->parent)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
if (PointerIsValid(target->name) && strcmp(target->name, name) == 0)
break;
}
if (!PointerIsValid(target))
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist", name)));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
Fix misplacement of savepointLevel test, per report from Chris K-L.
2004-08-03 17:57:26 +02:00
/* disallow crossing savepoint level boundaries */
if (target->savepointLevel != s->savepointLevel)
ereport(ERROR,
(errcode(ERRCODE_S_E_INVALID_SPECIFICATION),
Improve savepoint error messages Include the savepoint name in the error message and rephrase it a bit to match common style. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-18 02:29:27 +01:00
errmsg("savepoint \"%s\" does not exist within current savepoint level", name)));
Fix misplacement of savepointLevel test, per report from Chris K-L.
2004-08-03 17:57:26 +02:00
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
/*
* Mark "abort pending" all subtransactions up to the target
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* subtransaction. The actual aborts will happen when control gets to
* CommitTransactionCommand.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
xact = CurrentTransactionState;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
for (;;)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (xact == target)
break;
if (xact->blockState == TBLOCK_SUBINPROGRESS)
xact->blockState = TBLOCK_SUBABORT_PENDING;
else if (xact->blockState == TBLOCK_SUBABORT)
xact->blockState = TBLOCK_SUBABORT_END;
else
elog(FATAL, "RollbackToSavepoint: unexpected state %s",
BlockStateAsString(xact->blockState));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
xact = xact->parent;
Assert(PointerIsValid(xact));
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* And mark the target as "restart pending" */
if (xact->blockState == TBLOCK_SUBINPROGRESS)
xact->blockState = TBLOCK_SUBRESTART;
else if (xact->blockState == TBLOCK_SUBABORT)
xact->blockState = TBLOCK_SUBABORT_RESTART;
else
elog(FATAL, "RollbackToSavepoint: unexpected state %s",
BlockStateAsString(xact->blockState));
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
}
/*
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
* BeginInternalSubTransaction
Fix overly-strict sanity check in BeginInternalSubTransaction that made it fail when used in a deferred trigger. Bug goes back to 8.0; no doubt the reason it hadn't been noticed is that we've been discouraging use of user-defined constraint triggers. Per report from Frank van Vugt.
2007-05-30 23:01:39 +02:00
* This is the same as DefineSavepoint except it allows TBLOCK_STARTED,
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
* TBLOCK_IMPLICIT_INPROGRESS, TBLOCK_END, and TBLOCK_PREPARE states,
* and therefore it can safely be used in functions that might be called
* when not inside a BEGIN block or when running deferred triggers at
* COMMIT/PREPARE time. Also, it automatically does
* CommitTransactionCommand/StartTransactionCommand instead of expecting
* the caller to do it.
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
*/
void
Add some const decorations to prototypes Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
2017-10-31 15:34:31 +01:00
BeginInternalSubTransaction(const char *name)
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
{
TransactionState s = CurrentTransactionState;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for new subtransactions after that
* point. We might be able to make an exception for the type of
* subtransaction established by this function, which is typically used in
* contexts where we're going to release or roll back the subtransaction
* before proceeding further, so that no enduring change to the
* transaction state occurs. For now, however, we prohibit this case along
* with all the others.
*/
if (IsInParallelMode())
ereport(ERROR,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot start subtransactions during a parallel operation")));
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
switch (s->blockState)
{
case TBLOCK_STARTED:
case TBLOCK_INPROGRESS:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Fix overly-strict sanity check in BeginInternalSubTransaction that made it fail when used in a deferred trigger. Bug goes back to 8.0; no doubt the reason it hadn't been noticed is that we've been discouraging use of user-defined constraint triggers. Per report from Frank van Vugt.
2007-05-30 23:01:39 +02:00
case TBLOCK_END:
case TBLOCK_PREPARE:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_SUBINPROGRESS:
/* Normal subtransaction start */
PushTransaction();
s = CurrentTransactionState; /* changed by push */
Pgindent run for 8.0.
2004-08-29 07:07:03 +02:00
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Savepoint names, like the TransactionState block itself, live
* in TopTransactionContext.
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
*/
if (name)
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->name = MemoryContextStrdup(TopTransactionContext, name);
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_BEGIN:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_SUBBEGIN:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_ABORT:
case TBLOCK_SUBABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
elog(FATAL, "BeginInternalSubTransaction: unexpected state %s",
BlockStateAsString(s->blockState));
break;
}
CommitTransactionCommand();
StartTransactionCommand();
}
/*
* ReleaseCurrentSubTransaction
*
* RELEASE (ie, commit) the innermost subtransaction, regardless of its
* savepoint name (if any).
* NB: do NOT use CommitTransactionCommand/StartTransactionCommand with this.
*/
void
ReleaseCurrentSubTransaction(void)
{
TransactionState s = CurrentTransactionState;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Workers synchronize transaction state at the beginning of each parallel
* operation, so we can't account for commit of subtransactions after that
* point. This should not happen anyway. Code calling this would
* typically have called BeginInternalSubTransaction() first, failing
* there.
*/
if (IsInParallelMode())
ereport(ERROR,
(errcode(ERRCODE_INVALID_TRANSACTION_STATE),
errmsg("cannot commit subtransactions during a parallel operation")));
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
if (s->blockState != TBLOCK_SUBINPROGRESS)
elog(ERROR, "ReleaseCurrentSubTransaction: unexpected state %s",
BlockStateAsString(s->blockState));
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
Assert(s->state == TRANS_INPROGRESS);
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
MemoryContextSwitchTo(CurTransactionContext);
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
CommitSubTransaction();
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
s = CurrentTransactionState; /* changed by pop */
Assert(s->state == TRANS_INPROGRESS);
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
}
/*
* RollbackAndReleaseCurrentSubTransaction
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
* ROLLBACK and RELEASE (ie, abort) the innermost subtransaction, regardless
* of its savepoint name (if any).
* NB: do NOT use CommitTransactionCommand/StartTransactionCommand with this.
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*/
void
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
RollbackAndReleaseCurrentSubTransaction(void)
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
{
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* Unlike ReleaseCurrentSubTransaction(), this is nominally permitted
code: replace 'master' with 'leader' where appropriate. Leader already is the more widely used terminology, but a few places didn't get the message. Author: Andres Freund Reviewed-By: David Steele Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue@alap3.anarazel.de
2020-06-14 23:22:47 +02:00
* during parallel operations. That's because we may be in the leader,
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
* recovering from an error thrown while we were in parallel mode. We
* won't reach here in a worker, because BeginInternalSubTransaction()
* will have failed.
*/
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
switch (s->blockState)
{
/* Must be in a subtransaction */
case TBLOCK_SUBINPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/* These cases are invalid. */
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_DEFAULT:
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
case TBLOCK_INPROGRESS:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT:
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
plpgsql does exceptions. There are still some things that need refinement; in particular I fear that the recognized set of error condition names probably has little in common with what Oracle recognizes. But it's a start.
2004-07-31 09:39:21 +02:00
elog(FATAL, "RollbackAndReleaseCurrentSubTransaction: unexpected state %s",
BlockStateAsString(s->blockState));
break;
}
/*
* Abort the current subtransaction, if needed.
*/
if (s->blockState == TBLOCK_SUBINPROGRESS)
AbortSubTransaction();
/* And clean it up, too */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CleanupSubTransaction();
s = CurrentTransactionState; /* changed by pop */
Remove AssertArg and AssertState These don't offer anything over plain Assert, and their usage had already been declared obsolescent. Author: Nathan Bossart <nathandbossart@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Discussion: https://www.postgresql.org/message-id/20221009210148.GA900071@nathanxps13
2022-10-28 09:19:06 +02:00
Assert(s->blockState == TBLOCK_SUBINPROGRESS ||
Pre-beta mechanical code beautification. Run pgindent, pgperltidy, and reformat-dat-files. This set of diffs is a bit larger than typical. We've updated to pg_bsd_indent 2.1.2, which properly indents variable declarations that have multi-line initialization expressions (the continuation lines are now indented one tab stop). We've also updated to perltidy version 20230309 and changed some of its settings, which reduces its desire to add whitespace to lines to make assignments etc. line up. Going forward, that should make for fewer random-seeming changes to existing code. Discussion: https://postgr.es/m/20230428092545.qfb3y5wcu4cm75ur@alvherre.pgsql
2023-05-19 23:24:48 +02:00
s->blockState == TBLOCK_INPROGRESS ||
s->blockState == TBLOCK_IMPLICIT_INPROGRESS ||
s->blockState == TBLOCK_STARTED);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Remove unneeded dash blocks around function start comments.
2003-03-14 23:40:31 +01:00
/*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* AbortOutOfAnyTransaction
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
*
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
* This routine is provided for error recovery purposes. It aborts any
* active transaction or transaction block, leaving the system in a known
* idle state.
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
*/
void
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
AbortOutOfAnyTransaction(void)
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
{
TransactionState s = CurrentTransactionState;
Handle elog(FATAL) during ROLLBACK more robustly. Stress testing by Andreas Seltenreich disclosed longstanding problems that occur if a FATAL exit (e.g. due to receipt of SIGTERM) occurs while we are trying to execute a ROLLBACK of an already-failed transaction. In such a case, xact.c is in TBLOCK_ABORT state, so that AbortOutOfAnyTransaction would skip AbortTransaction and go straight to CleanupTransaction. This led to an assert failure in an assert-enabled build (due to the ROLLBACK's portal still having a cleanup hook) or without assertions, to a FATAL exit complaining about "cannot drop active portal". The latter's not disastrous, perhaps, but it's messy enough to want to improve it. We don't really want to run all of AbortTransaction in this code path. The minimum required to clean up the open portal safely is to do AtAbort_Memory and AtAbort_Portals. It seems like a good idea to do AtAbort_Memory unconditionally, to be entirely sure that we are starting with a safe CurrentMemoryContext. That means that if the main loop in AbortOutOfAnyTransaction does nothing, we need an extra step at the bottom to restore CurrentMemoryContext = TopMemoryContext, which I chose to do by invoking AtCleanup_Memory. This'll result in calling AtCleanup_Memory twice in many of the paths through this function, but that seems harmless and reasonably inexpensive. The original motivation for the assertion in AtCleanup_Portals was that we wanted to be sure that any user-defined code executed as a consequence of the cleanup hook runs during AbortTransaction not CleanupTransaction. That still seems like a valid concern, and now that we've seen one case of the assertion firing --- which means that exactly that would have happened in a production build --- let's replace the Assert with a runtime check. If we see the cleanup hook still set, we'll emit a WARNING and just drop the hook unexecuted. This has been like this a long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/877ey7bmun.fsf@ansel.ydns.eu
2017-08-14 21:43:20 +02:00
/* Ensure we're not running in a doomed memory context */
AtAbort_Memory();
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
/*
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
* Get out of any transaction or nested transaction
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
do
{
switch (s->blockState)
{
case TBLOCK_DEFAULT:
Teach AbortOutOfAnyTransaction to clean up partially-started transactions. AbortOutOfAnyTransaction failed to do anything if the state it saw on entry corresponded to failing partway through StartTransaction. I fixed AbortCurrentTransaction to cope with that case way back in commit 60b2444cc3ba037630c9b940c3c9ef01b954b87b, but evidently overlooked that AbortOutOfAnyTransaction should do likewise. Back-patch to all supported branches. It's not clear that this omission has any more-than-cosmetic consequences, but it's also not clear that it doesn't, so back-patching seems the least risky choice.
2012-05-29 05:57:06 +02:00
if (s->state == TRANS_DEFAULT)
{
/* Not in a transaction, do nothing */
}
else
{
/*
* We can get here after an error during transaction start
* (state will be TRANS_START). Need to clean up the
* incompletely started transaction. First, adjust the
* low-level state to suppress warning message from
* AbortTransaction.
*/
if (s->state == TRANS_START)
s->state = TRANS_INPROGRESS;
AbortTransaction();
CleanupTransaction();
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
break;
case TBLOCK_STARTED:
case TBLOCK_BEGIN:
case TBLOCK_INPROGRESS:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_END:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_PENDING:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* In a transaction, so clean up */
AbortTransaction();
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
break;
case TBLOCK_ABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
Handle elog(FATAL) during ROLLBACK more robustly. Stress testing by Andreas Seltenreich disclosed longstanding problems that occur if a FATAL exit (e.g. due to receipt of SIGTERM) occurs while we are trying to execute a ROLLBACK of an already-failed transaction. In such a case, xact.c is in TBLOCK_ABORT state, so that AbortOutOfAnyTransaction would skip AbortTransaction and go straight to CleanupTransaction. This led to an assert failure in an assert-enabled build (due to the ROLLBACK's portal still having a cleanup hook) or without assertions, to a FATAL exit complaining about "cannot drop active portal". The latter's not disastrous, perhaps, but it's messy enough to want to improve it. We don't really want to run all of AbortTransaction in this code path. The minimum required to clean up the open portal safely is to do AtAbort_Memory and AtAbort_Portals. It seems like a good idea to do AtAbort_Memory unconditionally, to be entirely sure that we are starting with a safe CurrentMemoryContext. That means that if the main loop in AbortOutOfAnyTransaction does nothing, we need an extra step at the bottom to restore CurrentMemoryContext = TopMemoryContext, which I chose to do by invoking AtCleanup_Memory. This'll result in calling AtCleanup_Memory twice in many of the paths through this function, but that seems harmless and reasonably inexpensive. The original motivation for the assertion in AtCleanup_Portals was that we wanted to be sure that any user-defined code executed as a consequence of the cleanup hook runs during AbortTransaction not CleanupTransaction. That still seems like a valid concern, and now that we've seen one case of the assertion firing --- which means that exactly that would have happened in a production build --- let's replace the Assert with a runtime check. If we see the cleanup hook still set, we'll emit a WARNING and just drop the hook unexecuted. This has been like this a long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/877ey7bmun.fsf@ansel.ydns.eu
2017-08-14 21:43:20 +02:00
/*
* AbortTransaction is already done, still need Cleanup.
* However, if we failed partway through running ROLLBACK,
* there will be an active portal running that command, which
* we need to shut down before doing CleanupTransaction.
*/
AtAbort_Portals();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CleanupTransaction();
s->blockState = TBLOCK_DEFAULT;
break;
Pgindent run for 8.0.
2004-08-29 07:07:03 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* In a subtransaction, so clean it up and abort parent too
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
AbortSubTransaction();
CleanupSubTransaction();
s = CurrentTransactionState; /* changed by pop */
break;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_END:
case TBLOCK_SUBABORT_RESTART:
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
/* As above, but AbortSubTransaction already done */
Handle elog(FATAL) during ROLLBACK more robustly. Stress testing by Andreas Seltenreich disclosed longstanding problems that occur if a FATAL exit (e.g. due to receipt of SIGTERM) occurs while we are trying to execute a ROLLBACK of an already-failed transaction. In such a case, xact.c is in TBLOCK_ABORT state, so that AbortOutOfAnyTransaction would skip AbortTransaction and go straight to CleanupTransaction. This led to an assert failure in an assert-enabled build (due to the ROLLBACK's portal still having a cleanup hook) or without assertions, to a FATAL exit complaining about "cannot drop active portal". The latter's not disastrous, perhaps, but it's messy enough to want to improve it. We don't really want to run all of AbortTransaction in this code path. The minimum required to clean up the open portal safely is to do AtAbort_Memory and AtAbort_Portals. It seems like a good idea to do AtAbort_Memory unconditionally, to be entirely sure that we are starting with a safe CurrentMemoryContext. That means that if the main loop in AbortOutOfAnyTransaction does nothing, we need an extra step at the bottom to restore CurrentMemoryContext = TopMemoryContext, which I chose to do by invoking AtCleanup_Memory. This'll result in calling AtCleanup_Memory twice in many of the paths through this function, but that seems harmless and reasonably inexpensive. The original motivation for the assertion in AtCleanup_Portals was that we wanted to be sure that any user-defined code executed as a consequence of the cleanup hook runs during AbortTransaction not CleanupTransaction. That still seems like a valid concern, and now that we've seen one case of the assertion firing --- which means that exactly that would have happened in a production build --- let's replace the Assert with a runtime check. If we see the cleanup hook still set, we'll emit a WARNING and just drop the hook unexecuted. This has been like this a long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/877ey7bmun.fsf@ansel.ydns.eu
2017-08-14 21:43:20 +02:00
if (s->curTransactionOwner)
{
/* As in TBLOCK_ABORT, might have a live portal to zap */
AtSubAbort_Portals(s->subTransactionId,
s->parent->subTransactionId,
s->curTransactionOwner,
s->parent->curTransactionOwner);
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CleanupSubTransaction();
s = CurrentTransactionState; /* changed by pop */
break;
}
} while (s->blockState != TBLOCK_DEFAULT);
pgindent run over code.
1999-05-25 18:15:34 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* Should be out of all subxacts now */
Assert(s->parent == NULL);
Handle elog(FATAL) during ROLLBACK more robustly. Stress testing by Andreas Seltenreich disclosed longstanding problems that occur if a FATAL exit (e.g. due to receipt of SIGTERM) occurs while we are trying to execute a ROLLBACK of an already-failed transaction. In such a case, xact.c is in TBLOCK_ABORT state, so that AbortOutOfAnyTransaction would skip AbortTransaction and go straight to CleanupTransaction. This led to an assert failure in an assert-enabled build (due to the ROLLBACK's portal still having a cleanup hook) or without assertions, to a FATAL exit complaining about "cannot drop active portal". The latter's not disastrous, perhaps, but it's messy enough to want to improve it. We don't really want to run all of AbortTransaction in this code path. The minimum required to clean up the open portal safely is to do AtAbort_Memory and AtAbort_Portals. It seems like a good idea to do AtAbort_Memory unconditionally, to be entirely sure that we are starting with a safe CurrentMemoryContext. That means that if the main loop in AbortOutOfAnyTransaction does nothing, we need an extra step at the bottom to restore CurrentMemoryContext = TopMemoryContext, which I chose to do by invoking AtCleanup_Memory. This'll result in calling AtCleanup_Memory twice in many of the paths through this function, but that seems harmless and reasonably inexpensive. The original motivation for the assertion in AtCleanup_Portals was that we wanted to be sure that any user-defined code executed as a consequence of the cleanup hook runs during AbortTransaction not CleanupTransaction. That still seems like a valid concern, and now that we've seen one case of the assertion firing --- which means that exactly that would have happened in a production build --- let's replace the Assert with a runtime check. If we see the cleanup hook still set, we'll emit a WARNING and just drop the hook unexecuted. This has been like this a long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/877ey7bmun.fsf@ansel.ydns.eu
2017-08-14 21:43:20 +02:00
/* If we didn't actually have anything to do, revert to TopMemoryContext */
AtCleanup_Memory();
Substantial rewrite of async.c to avoid problems with non-reentrant stdio and possibly other problems. Minor changes in xact.c and postgres.c's main loop to support new handling of async NOTIFY.
1998-10-06 04:40:09 +02:00
}
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
/*
* IsTransactionBlock --- are we within a transaction block?
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
bool
Fix AbortOutOfAnyTransaction logic to avoid notice about 'AbortTransaction and not in in-progress state' when client disconnects just after an error. Notice seems pretty harmless, so I'm not going to worry about back-patching this into 7.0.* ...
2000-10-24 22:06:39 +02:00
IsTransactionBlock(void)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
TransactionState s = CurrentTransactionState;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
if (s->blockState == TBLOCK_DEFAULT || s->blockState == TBLOCK_STARTED)
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return false;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return true;
}
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
/*
* IsTransactionOrTransactionBlock --- are we within either a transaction
* or a transaction block? (The backend is only really "idle" when this
* returns false.)
*
* This should match up with IsTransactionBlock and IsTransactionState.
*/
bool
IsTransactionOrTransactionBlock(void)
{
TransactionState s = CurrentTransactionState;
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
if (s->blockState == TBLOCK_DEFAULT)
Fix bad interaction between NOTIFY processing and V3 extended query protocol, per report from Igor Shevchenko. NOTIFY thought it could do its thing if transaction blockState is TBLOCK_DEFAULT, but in reality it had better check the low-level transaction state is TRANS_DEFAULT as well. Formerly it was not possible to wait for the client in a state where the first is true and the second is not ... but now we can have such a state. Minor cleanup in StartTransaction() as well.
2003-10-16 18:50:41 +02:00
return false;
return true;
}
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
/*
* TransactionBlockStatusCode - return status code to send in ReadyForQuery
*/
char
TransactionBlockStatusCode(void)
{
TransactionState s = CurrentTransactionState;
switch (s->blockState)
{
case TBLOCK_DEFAULT:
This is a cleanup patch for access/transam/xact.c. It only removes some #ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact that StartTransaction() has been executed. Alvaro Herrera
2004-04-05 05:11:39 +02:00
case TBLOCK_STARTED:
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return 'I'; /* idle --- not in transaction */
case TBLOCK_BEGIN:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBBEGIN:
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
case TBLOCK_INPROGRESS:
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_END:
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
case TBLOCK_SUBCOMMIT:
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return 'T'; /* in transaction */
case TBLOCK_ABORT:
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
case TBLOCK_SUBABORT_END:
case TBLOCK_ABORT_PENDING:
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
case TBLOCK_SUBABORT_RESTART:
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return 'E'; /* in failed transaction */
}
/* should never get here */
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
elog(FATAL, "invalid transaction block state: %s",
BlockStateAsString(s->blockState));
Add transaction status field to ReadyForQuery messages, and make room for tableID/columnID in RowDescription. (The latter isn't really implemented yet though --- the backend always sends zeroes, and libpq just throws away the data.)
2003-04-26 22:23:00 +02:00
return 0; /* keep compiler quiet */
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* IsSubTransaction
*/
bool
IsSubTransaction(void)
{
TransactionState s = CurrentTransactionState;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
if (s->nestingLevel >= 2)
return true;
return false;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
/*
* StartSubTransaction
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* If you're wondering why this is separate from PushTransaction: it's because
* we can't conveniently do this stuff right inside DefineSavepoint. The
* SAVEPOINT utility command will be executed inside a Portal, and if we
* muck with CurrentMemoryContext or CurrentResourceOwner then exit from
* the Portal will undo those settings. So we make DefineSavepoint just
* push a dummy transaction block, and when control returns to the main
* idle loop, CommitTransactionCommand will be called, and we'll come here
* to finish starting the subtransaction.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
static void
StartSubTransaction(void)
{
TransactionState s = CurrentTransactionState;
if (s->state != TRANS_DEFAULT)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "StartSubTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->state = TRANS_START;
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Initialize subsystems for new subtransaction
*
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
* must initialize resource-management stuff first
*/
AtSubStart_Memory();
AtSubStart_ResourceOwner();
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
AfterTriggerBeginSubXact();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->state = TRANS_INPROGRESS;
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
/*
* Call start-of-subxact callbacks
*/
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CallSubXactCallbacks(SUBXACT_EVENT_START_SUB, s->subTransactionId,
s->parent->subTransactionId);
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
ShowTransactionState("StartSubTransaction");
}
/*
* CommitSubTransaction
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* The caller has to make sure to always reassign CurrentTransactionState
* if it has a local pointer to it after calling this function.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
static void
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
CommitSubTransaction(void)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
{
TransactionState s = CurrentTransactionState;
ShowTransactionState("CommitSubTransaction");
if (s->state != TRANS_INPROGRESS)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "CommitSubTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Invent pre-commit/pre-prepare/pre-subcommit events for xact callbacks. Currently it's only possible for loadable modules to get control during post-commit cleanup of a transaction. That doesn't work too well if they want to do something that could throw an error; for example, an FDW might need to issue a remote commit, which could well fail. To improve matters, extend the existing APIs for XactCallback and SubXactCallback functions to provide new pre-commit events for this purpose. The release notes will need to mention that existing callback functions should be checked to make sure they don't do something unwanted when one of the new event types occurs. In the examples within our source tree, contrib/sepgsql was fine but plpgsql had been a bit too cute.
2013-02-15 02:35:08 +01:00
/* Pre-commit processing goes here */
CallSubXactCallbacks(SUBXACT_EVENT_PRE_COMMIT_SUB, s->subTransactionId,
s->parent->subTransactionId);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/* If in parallel mode, clean up workers and exit parallel mode. */
if (IsInParallelMode())
{
AtEOSubXact_Parallel(true, s->subTransactionId);
s->parallelModeLevel = 0;
}
Invent pre-commit/pre-prepare/pre-subcommit events for xact callbacks. Currently it's only possible for loadable modules to get control during post-commit cleanup of a transaction. That doesn't work too well if they want to do something that could throw an error; for example, an FDW might need to issue a remote commit, which could well fail. To improve matters, extend the existing APIs for XactCallback and SubXactCallback functions to provide new pre-commit events for this purpose. The release notes will need to mention that existing callback functions should be checked to make sure they don't do something unwanted when one of the new event types occurs. In the examples within our source tree, contrib/sepgsql was fine but plpgsql had been a bit too cute.
2013-02-15 02:35:08 +01:00
/* Do the actual "commit", such as it is */
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
s->state = TRANS_COMMIT;
Redesign query-snapshot timing so that volatile functions in READ COMMITTED mode see a fresh snapshot for each command in the function, rather than using the latest interactive command's snapshot. Also, suppress fresh snapshots as well as CommandCounterIncrement inside STABLE and IMMUTABLE functions, instead using the snapshot taken for the most closely nested regular query. (This behavior is only sane for read-only functions, so the patch also enforces that such functions contain only SELECT commands.) As per my proposal of 6-Sep-2004; I note that I floated essentially the same proposal on 19-Jun-2002, but that discussion tailed off without any action. Since 8.0 seems like the right place to be taking possibly nontrivial backwards compatibility hits, let's get it done now.
2004-09-13 22:10:13 +02:00
/* Must CCI to ensure commands of subtransaction are seen as done */
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
CommandCounterIncrement();
Redesign query-snapshot timing so that volatile functions in READ COMMITTED mode see a fresh snapshot for each command in the function, rather than using the latest interactive command's snapshot. Also, suppress fresh snapshots as well as CommandCounterIncrement inside STABLE and IMMUTABLE functions, instead using the snapshot taken for the most closely nested regular query. (This behavior is only sane for read-only functions, so the patch also enforces that such functions contain only SELECT commands.) As per my proposal of 6-Sep-2004; I note that I floated essentially the same proposal on 19-Jun-2002, but that discussion tailed off without any action. Since 8.0 seems like the right place to be taking possibly nontrivial backwards compatibility hits, let's get it done now.
2004-09-13 22:10:13 +02:00
Rework subtransaction commit protocol for hot standby. This patch eliminates the marking of subtransactions as SUBCOMMITTED in pg_clog during their commit; instead they remain in-progress until main transaction commit. At main transaction commit, the commit protocol is atomic-by-page instead of one transaction at a time. To avoid a race condition with some subtransactions appearing committed before others in the case where they span more than one pg_clog page, we conserve the logic that marks them subcommitted before marking the parent committed. Simon Riggs with minor help from me
2008-10-20 21:18:18 +02:00
/*
* Prior to 8.4 we marked subcommit in clog at this point. We now only
* perform that step, if required, as part of the atomic update of the
* whole transaction tree at top level commit or abort.
*/
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* Post-commit cleanup */
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
AtSubCommit_childXids();
Fire non-deferred AFTER triggers immediately upon query completion, rather than when returning to the idle loop. This makes no particular difference for interactively-issued queries, but it makes a big difference for queries issued within functions: trigger execution now occurs before the calling function is allowed to proceed. This responds to numerous complaints about nonintuitive behavior of foreign key checking, such as http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and appears to be required by the SQL99 spec. Also take the opportunity to simplify the data structures used for the pending-trigger list, rename them for more clarity, and squeeze out a bit of space.
2004-09-10 20:40:09 +02:00
AfterTriggerEndSubXact(true);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtSubCommit_Portals(s->subTransactionId,
s->parent->subTransactionId,
Fix Portal snapshot tracking to handle subtransactions properly. Commit 84f5c2908 forgot to consider the possibility that EnsurePortalSnapshotExists could run inside a subtransaction with lifespan shorter than the Portal's. In that case, the new active snapshot would be popped at the end of the subtransaction, leaving a dangling pointer in the Portal, with mayhem ensuing. To fix, make sure the ActiveSnapshot stack entry is marked with the same subtransaction nesting level as the associated Portal. It's certainly safe to do so since we won't be here at all unless the stack is empty; hence we can't create an out-of-order stack. Let's also apply this logic in the case where PortalRunUtility sets portalSnapshot, just to be sure that path can't cause similar problems. It's slightly less clear that that path can't create an out-of-order stack, so add an assertion guarding it. Report and patch by Bertrand Drouvot (with kibitzing by me). Back-patch to v11, like the previous commit. Discussion: https://postgr.es/m/ff82b8c5-77f4-3fe7-6028-fcf3303e82dd@amazon.com
2021-10-01 17:10:12 +02:00
s->parent->nestingLevel,
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
s->parent->curTransactionOwner);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtEOSubXact_LargeObject(true, s->subTransactionId,
s->parent->subTransactionId);
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
AtSubCommit_Notify();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CallSubXactCallbacks(SUBXACT_EVENT_COMMIT_SUB, s->subTransactionId,
s->parent->subTransactionId);
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
true, false);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtEOSubXact_RelationCache(true, s->subTransactionId,
s->parent->subTransactionId);
Some mop-up work for savepoints (nested transactions). Store a small number of active subtransaction XIDs in each backend's PGPROC entry, and use this to avoid expensive probes into pg_subtrans during TransactionIdIsInProgress. Extend EOXactCallback API to allow add-on modules to get control at subxact start/end. (This is deliberately not compatible with the former API, since any uses of that API probably need manual review anyway.) Add basic reference documentation for SAVEPOINT and related commands. Minor other cleanups to check off some of the open issues for subtransactions. Alvaro Herrera and Tom Lane.
2004-08-01 19:32:22 +02:00
AtEOSubXact_Inval(true);
Fix incorrect ordering of smgr cleanup relative to buffer pin cleanup during transaction abort. Add a regression test case to catch related mistakes in future. Alvaro Herrera and Tom Lane.
2004-09-06 19:56:33 +02:00
AtSubCommit_smgr();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* The only lock we actually release here is the subtransaction XID lock.
*/
CurrentResourceOwner = s->curTransactionOwner;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
XactLockTableDelete(XidFromFullTransactionId(s->fullTransactionId));
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
/*
* Other locks should get transferred to their parent resource owner.
*/
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_LOCKS,
Partially revoke attempt to improve performance with many savepoints. Maintain difference between subtransaction release and commit introduced by earlier patch.
2011-09-07 13:11:26 +02:00
true, false);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_AFTER_LOCKS,
true, false);
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
AtEOXact_GUC(true, s->gucNestLevel);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtEOSubXact_SPI(true, s->subTransactionId);
AtEOSubXact_on_commit_actions(true, s->subTransactionId,
s->parent->subTransactionId);
AtEOSubXact_Namespace(true, s->subTransactionId,
s->parent->subTransactionId);
AtEOSubXact_Files(true, s->subTransactionId,
s->parent->subTransactionId);
Fix dynahash.c to suppress hash bucket splits while a hash_seq_search() scan is in progress on the same hashtable. This seems the least invasive way to fix the recently-recognized problem that a split could cause the scan to visit entries twice or (with much lower probability) miss them entirely. The only field-reported problem caused by this is the "failed to re-find shared lock object" PANIC in COMMIT PREPARED reported by Michel Dorochevsky, which was caused by multiply visited entries. However, it seems certain that mdsync() is vulnerable to missing required fsync's due to missed entries, and I am fearful that RelationCacheInitializePhase2() might be at risk as well. Because of that and the generalized hazard presented by this bug, back-patch all the supported branches. Along the way, fix pg_prepared_statement() and pg_cursor() to not assume that the hashtables they are examining will stay static between calls. This is risky regardless of the newly noted dynahash problem, because hash_seq_search() has never promised to cope with deletion of table entries other than the just-returned one. There may be no bug here because the only supported way to call these functions is via ExecMakeTableFunctionResult() which will cycle them to completion before doing anything very interesting, but it seems best to get rid of the assumption. This affects 8.2 and HEAD only, since those functions weren't there earlier.
2007-04-27 01:24:46 +02:00
AtEOSubXact_HashTables(true, s->nestingLevel);
Fix up pgstats counting of live and dead tuples to recognize that committed and aborted transactions have different effects; also teach it not to assume that prepared transactions are always committed. Along the way, simplify the pgstats API by tying counting directly to Relations; I cannot detect any redeeming social value in having stats pointers in HeapScanDesc and IndexScanDesc structures. And fix a few corner cases in which counts might be missed because the relation's pgstat_info pointer hadn't been set.
2007-05-27 05:50:39 +02:00
AtEOSubXact_PgStat(true, s->nestingLevel);
Improve snapshot manager by keeping explicit track of snapshots. There are two ways to track a snapshot: there's the "registered" list, which is used for arbitrary long-lived snapshots; and there's the "active stack", which is used for the snapshot that is considered "active" at any time. This also allows users of snapshots to stop worrying about snapshot memory allocation and freeing, and about using PG_TRY blocks around ActiveSnapshot assignment. This is all done automatically now. As a consequence, this allows us to reset MyProc->xmin when there are no more snapshots registered in the current backend, reducing the impact that long-running transactions have on VACUUM.
2008-05-12 22:02:02 +02:00
AtSubCommit_Snapshot(s->nestingLevel);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
/*
* We need to restore the upper transaction's read-only state, in case the
* upper is read-write while the child is read-only; GUC will incorrectly
* think it should leave the child state in place.
*/
XactReadOnly = s->prevXactReadOnly;
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
CurrentResourceOwner = s->parent->curTransactionOwner;
CurTransactionResourceOwner = s->parent->curTransactionOwner;
Revise ResourceOwner code to avoid accumulating ResourceOwner objects for every command executed within a transaction. For long transactions this was a significant memory leak. Instead, we can delete a portal's or subtransaction's ResourceOwner immediately, if we physically transfer the information about its locks up to the parent owner. This does not fully solve the leak problem; we need to do something about counting multiple acquisitions of the same lock in order to fix it. But it's a necessary step along the way.
2004-08-25 20:43:43 +02:00
ResourceOwnerDelete(s->curTransactionOwner);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
s->curTransactionOwner = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
AtSubCommit_Memory();
s->state = TRANS_DEFAULT;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
PopTransaction();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
/*
* AbortSubTransaction
*/
static void
AbortSubTransaction(void)
{
TransactionState s = CurrentTransactionState;
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/* Prevent cancel/die interrupt while cleaning up */
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
HOLD_INTERRUPTS();
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
/* Make sure we have a valid memory context and resource owner */
AtSubAbort_Memory();
AtSubAbort_ResourceOwner();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* Release any LW locks we might be holding as quickly as possible.
* (Regular locks, however, must be held till we finish aborting.)
* Releasing LW locks is critical since we might try to grab them again
* while cleaning up!
*
* FIXME This may be incorrect --- Are there some locks we should keep?
* Buffer locks, for example? I don't think so but I'm not sure.
*/
LWLockReleaseAll();
Provide much better wait information in pg_stat_activity. When a process is waiting for a heavyweight lock, we will now indicate the type of heavyweight lock for which it is waiting. Also, you can now see when a process is waiting for a lightweight lock - in which case we will indicate the individual lock name or the tranche, as appropriate - or for a buffer pin. Amit Kapila, Ildus Kurbangaliev, reviewed by me. Lots of helpful discussion and suggestions by many others, including Alexander Korotkov, Vladimir Borodin, and many others.
2016-03-10 18:44:09 +01:00
pgstat_report_wait_end();
pgstat_progress_end_command();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
UnlockBuffers();
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
/* Reset WAL record construction state */
XLogResetInsertion();
Cancel CV sleep during subtransaction abort. Generally, error recovery paths that need to do things like LWLockReleaseAll and pgstat_report_wait_end also need to call ConditionVariableCancelSleep, but AbortSubTransaction was missed. Since subtransaction abort might destroy up the DSM segment that contains the ConditionVariable stored in cv_sleep_target, this can result in a crash for anything using condition variables. Reported and diagnosed by Andres Freund. Discussion: http://postgr.es/m/20171221110048.rxk6464azzl5t2fi@alap3.anarazel.de
2017-12-21 15:09:04 +01:00
/* Cancel condition variable sleep */
ConditionVariableCancelSleep();
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
/*
* Also clean up any open wait for lock, since the lock manager will choke
* if we try to wait for another lock before doing this.
*/
Tighten up error recovery for fast-path locking. The previous code could cause a backend crash after BEGIN; SAVEPOINT a; LOCK TABLE foo (interrupted by ^C or statement timeout); ROLLBACK TO SAVEPOINT a; LOCK TABLE foo, and might have leaked strong-lock counts in other situations. Report by Zoltán Böszörményi; patch review by Jeff Davis.
2012-04-18 17:17:30 +02:00
LockErrorCleanup();
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
/*
* If any timeout events are still active, make sure the timeout interrupt
* is scheduled. This covers possible loss of a timeout interrupt due to
* longjmp'ing out of the SIGINT handler (see notes in handle_sig_alarm).
* We delay this till after LockErrorCleanup so that we don't uselessly
* reschedule lock or deadlock check timeouts.
*/
reschedule_timeouts();
/*
* Re-enable signals, in case we got here by longjmp'ing out of a signal
* handler. We do this fairly early in the sequence so that the timeout
* infrastructure will be functional if needed while aborting.
*/
Retire PG_SETMASK() macro. In the 90s we needed to deal with computers that still had the pre-standard signal masking APIs. That hasn't been relevant for a very long time on Unix systems, and c94ae9d8 got rid of a remaining dependency in our Windows porting code. PG_SETMASK didn't expose save/restore functionality, so we'd already started using sigprocmask() directly in places, creating the visual distraction of having two ways to spell it. It's not part of the API that extensions are expected to be using (but if they are, the change will be trivial). It seems like a good time to drop the old macro and just call the standard POSIX function. Reviewed-by: Nathan Bossart <nathandbossart@gmail.com> Discussion: https://postgr.es/m/CA%2BhUKG%2BKfQgrhHP2DLTohX1WwubaCBHmTzGnAEDPZ-Gug-Xskg%40mail.gmail.com
2023-02-02 22:34:56 +01:00
sigprocmask(SIG_SETMASK, &UnBlockSig, NULL);
Fix assorted race conditions in the new timeout infrastructure. Prevent handle_sig_alarm from losing control partway through due to a query cancel (either an asynchronous SIGINT, or a cancel triggered by one of the timeout handler functions). That would at least result in failure to schedule any required future interrupt, and might result in actual corruption of timeout.c's data structures, if the interrupt happened while we were updating those. We could still lose control if an asynchronous SIGINT arrives just as the function is entered. This wouldn't break any data structures, but it would have the same effect as if the SIGALRM interrupt had been silently lost: we'd not fire any currently-due handlers, nor schedule any new interrupt. To forestall that scenario, forcibly reschedule any pending timer interrupt during AbortTransaction and AbortSubTransaction. We can avoid any extra kernel call in most cases by not doing that until we've allowed LockErrorCleanup to kill the DEADLOCK_TIMEOUT and LOCK_TIMEOUT events. Another hazard is that some platforms (at least Linux and *BSD) block a signal before calling its handler and then unblock it on return. When we longjmp out of the handler, the unblock doesn't happen, and the signal is left blocked indefinitely. Again, we can fix that by forcibly unblocking signals during AbortTransaction and AbortSubTransaction. These latter two problems do not manifest when the longjmp reaches postgres.c, because the error recovery code there kills all pending timeout events anyway, and it uses sigsetjmp(..., 1) so that the appropriate signal mask is restored. So errors thrown outside any transaction should be OK already, and cleaning up in AbortTransaction and AbortSubTransaction should be enough to fix these issues. (We're assuming that any code that catches a query cancel error and doesn't re-throw it will do at least a subtransaction abort to clean up; but that was pretty much required already by other subsystems.) Lastly, ProcSleep should not clear the LOCK_TIMEOUT indicator flag when disabling that event: if a lock timeout interrupt happened after the lock was granted, the ensuing query cancel is still going to happen at the next CHECK_FOR_INTERRUPTS, and we want to report it as a lock timeout not a user cancel. Per reports from Dan Wood. Back-patch to 9.3 where the new timeout handling infrastructure was introduced. We may at some point decide to back-patch the signal unblocking changes further, but I'll desist from that until we hear actual field complaints about it.
2013-11-29 22:41:00 +01:00
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
* check the current transaction state
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
Several changes to reduce the probability of running out of memory during AbortTransaction, which would lead to recursion and eventual PANIC exit as illustrated in recent report from Jeff Davis. First, in xact.c create a special dedicated memory context for AbortTransaction to run in. This solves the problem as long as AbortTransaction doesn't need more than 32K (or whatever other size we create the context with). But in corner cases it might. Second, in trigger.c arrange to keep pending after-trigger event records in separate contexts that can be freed near the beginning of AbortTransaction, rather than having them persist until CleanupTransaction as before. Third, in portalmem.c arrange to free executor state data earlier as well. These two changes should result in backing off the out-of-memory condition before AbortTransaction needs any significant amount of memory, at least in typical cases such as memory overrun due to too many trigger events or too big an executor hash table. And all the same for subtransaction abort too, of course.
2006-11-23 02:14:59 +01:00
ShowTransactionState("AbortSubTransaction");
if (s->state != TRANS_INPROGRESS)
elog(WARNING, "AbortSubTransaction while in %s state",
TransStateAsString(s->state));
s->state = TRANS_ABORT;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Make standard maintenance operations (including VACUUM, ANALYZE, REINDEX, and CLUSTER) execute as the table owner rather than the calling user, using the same privilege-switching mechanism already used for SECURITY DEFINER functions. The purpose of this change is to ensure that user-defined functions used in index definitions cannot acquire the privileges of a superuser account that is performing routine maintenance. While a function used in an index is supposed to be IMMUTABLE and thus not able to do anything very interesting, there are several easy ways around that restriction; and even if we could plug them all, there would remain a risk of reading sensitive information and broadcasting it through a covert channel such as CPU usage. To prevent bypassing this security measure, execution of SET SESSION AUTHORIZATION and SET ROLE is now forbidden within a SECURITY DEFINER context. Thanks to Itagaki Takahiro for reporting this vulnerability. Security: CVE-2007-6600
2008-01-03 22:23:15 +01:00
/*
* Reset user ID which might have been changed transiently. (See notes in
* AbortTransaction.)
*/
Prevent indirect security attacks via changing session-local state within an allegedly immutable index function. It was previously recognized that we had to prevent such a function from executing SET/RESET ROLE/SESSION AUTHORIZATION, or it could trivially obtain the privileges of the session user. However, since there is in general no privilege checking for changes of session-local state, it is also possible for such a function to change settings in a way that might subvert later operations in the same session. Examples include changing search_path to cause an unexpected function to be called, or replacing an existing prepared statement with another one that will execute a function of the attacker's choosing. The present patch secures VACUUM, ANALYZE, and CREATE INDEX/REINDEX against these threats, which are the same places previously deemed to need protection against the SET ROLE issue. GUC changes are still allowed, since there are many useful cases for that, but we prevent security problems by forcing a rollback of any GUC change after completing the operation. Other cases are handled by throwing an error if any change is attempted; these include temp table creation, closing a cursor, and creating or deleting a prepared statement. (In 7.4, the infrastructure to roll back GUC changes doesn't exist, so we settle for rejecting changes of "search_path" in these contexts.) Original report and patch by Gurjeet Singh, additional analysis by Tom Lane. Security: CVE-2009-4136
2009-12-09 22:57:51 +01:00
SetUserIdAndSecContext(s->prevUser, s->prevSecContext);
Make standard maintenance operations (including VACUUM, ANALYZE, REINDEX, and CLUSTER) execute as the table owner rather than the calling user, using the same privilege-switching mechanism already used for SECURITY DEFINER functions. The purpose of this change is to ensure that user-defined functions used in index definitions cannot acquire the privileges of a superuser account that is performing routine maintenance. While a function used in an index is supposed to be IMMUTABLE and thus not able to do anything very interesting, there are several easy ways around that restriction; and even if we could plug them all, there would remain a risk of reading sensitive information and broadcasting it through a covert channel such as CPU usage. To prevent bypassing this security measure, execution of SET SESSION AUTHORIZATION and SET ROLE is now forbidden within a SECURITY DEFINER context. Thanks to Itagaki Takahiro for reporting this vulnerability. Security: CVE-2007-6600
2008-01-03 22:23:15 +01:00
Fix possible crash during FATAL exit from reindexing. index.c supposed that it could just use a PG_TRY block to clean up the state associated with an active REINDEX operation. However, that code doesn't run if we do a FATAL exit --- for example, due to a SIGTERM shutdown signal --- while the REINDEX is happening. And that state does get consulted during catalog accesses, which makes it problematic if we do any catalog accesses during shutdown --- for example, to clean up any temp tables created in the session. If this combination of circumstances occurred, we could find ourselves trying to access already-freed memory. In debug builds that'd fairly reliably cause an assertion failure. In production we might often get away with it, but with some bad luck it could cause a core dump. Another possible bad outcome is an erroneous conclusion that an index-to-be-accessed is being reindexed; but it looks like that would be unlikely to have any consequences worse than failing to drop temp tables right away. (They'd still get dropped by the next session that uses that temp schema.) To fix, get rid of the use of PG_TRY here, and instead hook into the transaction abort mechanisms to clean up reindex state. Per bug #16378 from Alexander Lakhin. This has been wrong for a very long time, so back-patch to all supported branches. Discussion: https://postgr.es/m/16378-7a70ca41b3ec2009@postgresql.org
2020-04-21 21:58:42 +02:00
/* Forget about any active REINDEX. */
ResetReindexState(s->nestingLevel);
Implement streaming mode in ReorderBuffer. Instead of serializing the transaction to disk after reaching the logical_decoding_work_mem limit in memory, we consume the changes we have in memory and invoke stream API methods added by commit 45fdc9738b. However, sometimes if we have incomplete toast or speculative insert we spill to the disk because we can't generate the complete tuple and stream. And, as soon as we get the complete tuple we stream the transaction including the serialized changes. We can do this incremental processing thanks to having assignments (associating subxact with toplevel xacts) in WAL right away, and thanks to logging the invalidation messages at each command end. These features are added by commits 0bead9af48 and c55040ccd0 respectively. Now that we can stream in-progress transactions, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We handle such failures by returning ERRCODE_TRANSACTION_ROLLBACK sqlerrcode from system table scan APIs to the backend or WALSender decoding a specific uncommitted transaction. The decoding logic on the receipt of such a sqlerrcode aborts the decoding of the current transaction and continue with the decoding of other transactions. We have ReorderBufferTXN pointer in each ReorderBufferChange by which we know which xact it belongs to. The output plugin can use this to decide which changes to discard in case of stream_abort_cb (e.g. when a subxact gets discarded). We also provide a new option via SQL APIs to fetch the changes being streamed. Author: Dilip Kumar, Tomas Vondra, Amit Kapila, Nikhil Sontakke Reviewed-by: Amit Kapila, Kuntal Ghosh, Ajin Cherian Tested-by: Neha Sharma, Mahendra Singh Thalor and Ajin Cherian Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2020-08-08 04:04:39 +02:00
/* Reset logical streaming state. */
ResetLogicalStreamingState();
Reset properly snapshot export state during transaction abort During a replication slot creation, an ERROR generated in the same transaction as the one creating a to-be-exported snapshot would have left the backend in an inconsistent state, as the associated static export snapshot state was not being reset on transaction abort, but only on the follow-up command received by the WAL sender that created this snapshot on replication slot creation. This would trigger inconsistency failures if this session tried to export again a snapshot, like during the creation of a replication slot. Note that a snapshot export cannot happen in a transaction block, so there is no need to worry resetting this state for subtransaction aborts. Also, this inconsistent state would very unlikely show up to users. For example, one case where this could happen is an out-of-memory error when building the initial snapshot to-be-exported. Dilip found this problem while poking at a different patch, that caused an error in this code path for reasons unrelated to HEAD. Author: Dilip Kumar Reviewed-by: Michael Paquier, Zhihong Yu Discussion: https://postgr.es/m/CAFiTN-s0zA1Kj0ozGHwkYkHwa5U0zUE94RSc_g81WrpcETB5=w@mail.gmail.com Backpatch-through: 9.6
2021-10-18 04:55:42 +02:00
/*
* No need for SnapBuildResetExportedSnapshotState() here, snapshot
* exports are not supported in subtransactions.
*/
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/* Exit from parallel mode, if necessary. */
if (IsInParallelMode())
{
AtEOSubXact_Parallel(false, s->subTransactionId);
s->parallelModeLevel = 0;
}
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* We can skip all this stuff if the subxact failed before creating a
* ResourceOwner...
*/
if (s->curTransactionOwner)
{
AfterTriggerEndSubXact(false);
AtSubAbort_Portals(s->subTransactionId,
s->parent->subTransactionId,
Fix subtransaction cleanup after an outer-subtransaction portal fails. Formerly, we treated only portals created in the current subtransaction as having failed during subtransaction abort. However, if the error occurred while running a portal created in an outer subtransaction (ie, a cursor declared before the last savepoint), that has to be considered broken too. To allow reliable detection of which ones those are, add a bookkeeping field to struct Portal that tracks the innermost subtransaction in which each portal has actually been executed. (Without this, we'd end up failing portals containing functions that had called the subtransaction, thereby breaking plpgsql exception blocks completely.) In addition, when we fail an outer-subtransaction Portal, transfer its resources into the subtransaction's resource owner, so that they're released early in cleanup of the subxact. This fixes a problem reported by Jim Nasby in which a function executed in an outer-subtransaction cursor could cause an Assert failure or crash by referencing a relation created within the inner subtransaction. The proximate cause of the Assert failure is that AtEOSubXact_RelationCache assumed it could blow away a relcache entry without first checking that the entry had zero refcount. That was a bad idea on its own terms, so add such a check there, and to the similar coding in AtEOXact_RelationCache. This provides an independent safety measure in case there are still ways to provoke the situation despite the Portal-level changes. This has been broken since subtransactions were invented, so back-patch to all supported branches. Tom Lane and Michael Paquier
2015-09-04 19:36:49 +02:00
s->curTransactionOwner,
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->parent->curTransactionOwner);
AtEOSubXact_LargeObject(false, s->subTransactionId,
s->parent->subTransactionId);
AtSubAbort_Notify();
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
/* Advertise the fact that we aborted in pg_xact. */
Replace the former method of determining snapshot xmax --- to wit, calling ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid" variable that is updated during transaction commit or abort. Since latestCompletedXid is written only in places that had to lock ProcArrayLock exclusively anyway, and is read only in places that had to lock ProcArrayLock shared anyway, it adds no new locking requirements to the system despite being cluster-wide. Moreover, removing ReadNewTransactionId from snapshot acquisition eliminates the need to take both XidGenLock and ProcArrayLock at the same time. Since XidGenLock is sometimes held across I/O this can be a significant win. Some preliminary benchmarking suggested that this patch has no effect on average throughput but can significantly improve the worst-case transaction times seen in pgbench. Concept by Florian Pflug, implementation by Tom Lane.
2007-09-08 22:31:15 +02:00
(void) RecordTransactionAbort(true);
Implement lazy XID allocation: transactions that do not modify any database rows will normally never obtain an XID at all. We already did things this way for subtransactions, but this patch extends the concept to top-level transactions. In applications where there are lots of short read-only transactions, this should improve performance noticeably; not so much from removal of the actual XID-assignments, as from reduction of overhead that's driven by the rate of XID consumption. We add a concept of a "virtual transaction ID" so that active transactions can be uniquely identified even if they don't have a regular XID. This is a much lighter-weight concept: uniqueness of VXIDs is only guaranteed over the short term, and no on-disk record is made about them. Florian Pflug, with some editorialization by Tom.
2007-09-05 20:10:48 +02:00
/* Post-abort cleanup */
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
Add some marginal tweaks to eliminate memory leakages associated with subtransactions. Trivial subxacts (such as a plpgsql exception block containing no database access) now demonstrably leak zero bytes.
2004-09-16 22:17:49 +02:00
AtSubAbort_childXids();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CallSubXactCallbacks(SUBXACT_EVENT_ABORT_SUB, s->subTransactionId,
s->parent->subTransactionId);
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_BEFORE_LOCKS,
false, false);
AtEOSubXact_RelationCache(false, s->subTransactionId,
s->parent->subTransactionId);
AtEOSubXact_Inval(false);
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_LOCKS,
false, false);
ResourceOwnerRelease(s->curTransactionOwner,
RESOURCE_RELEASE_AFTER_LOCKS,
false, false);
During transaction cleanup, release locks before deleting files. There's no need to hold onto the locks until the files are needed, and by doing it this way, we reduce the impact on other backends who may be awaiting locks we hold. Noah Misch
2012-06-14 16:19:33 +02:00
AtSubAbort_smgr();
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
AtEOXact_GUC(false, s->gucNestLevel);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtEOSubXact_SPI(false, s->subTransactionId);
AtEOSubXact_on_commit_actions(false, s->subTransactionId,
s->parent->subTransactionId);
AtEOSubXact_Namespace(false, s->subTransactionId,
s->parent->subTransactionId);
AtEOSubXact_Files(false, s->subTransactionId,
s->parent->subTransactionId);
Fix dynahash.c to suppress hash bucket splits while a hash_seq_search() scan is in progress on the same hashtable. This seems the least invasive way to fix the recently-recognized problem that a split could cause the scan to visit entries twice or (with much lower probability) miss them entirely. The only field-reported problem caused by this is the "failed to re-find shared lock object" PANIC in COMMIT PREPARED reported by Michel Dorochevsky, which was caused by multiply visited entries. However, it seems certain that mdsync() is vulnerable to missing required fsync's due to missed entries, and I am fearful that RelationCacheInitializePhase2() might be at risk as well. Because of that and the generalized hazard presented by this bug, back-patch all the supported branches. Along the way, fix pg_prepared_statement() and pg_cursor() to not assume that the hashtables they are examining will stay static between calls. This is risky regardless of the newly noted dynahash problem, because hash_seq_search() has never promised to cope with deletion of table entries other than the just-returned one. There may be no bug here because the only supported way to call these functions is via ExecMakeTableFunctionResult() which will cycle them to completion before doing anything very interesting, but it seems best to get rid of the assumption. This affects 8.2 and HEAD only, since those functions weren't there earlier.
2007-04-27 01:24:46 +02:00
AtEOSubXact_HashTables(false, s->nestingLevel);
Fix up pgstats counting of live and dead tuples to recognize that committed and aborted transactions have different effects; also teach it not to assume that prepared transactions are always committed. Along the way, simplify the pgstats API by tying counting directly to Relations; I cannot detect any redeeming social value in having stats pointers in HeapScanDesc and IndexScanDesc structures. And fix a few corner cases in which counts might be missed because the relation's pgstat_info pointer hadn't been set.
2007-05-27 05:50:39 +02:00
AtEOSubXact_PgStat(false, s->nestingLevel);
Improve snapshot manager by keeping explicit track of snapshots. There are two ways to track a snapshot: there's the "registered" list, which is used for arbitrary long-lived snapshots; and there's the "active stack", which is used for the snapshot that is considered "active" at any time. This also allows users of snapshots to stop worrying about snapshot memory allocation and freeing, and about using PG_TRY blocks around ActiveSnapshot assignment. This is all done automatically now. As a consequence, this allows us to reset MyProc->xmin when there are no more snapshots registered in the current backend, reducing the impact that long-running transactions have on VACUUM.
2008-05-12 22:02:02 +02:00
AtSubAbort_Snapshot(s->nestingLevel);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
/*
* Restore the upper transaction's read-only state, too. This should be
* redundant with GUC's cleanup but we may as well do it for consistency
* with the commit case.
*/
XactReadOnly = s->prevXactReadOnly;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
RESUME_INTERRUPTS();
}
/*
* CleanupSubTransaction
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
*
* The caller has to make sure to always reassign CurrentTransactionState
* if it has a local pointer to it after calling this function.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
static void
CleanupSubTransaction(void)
{
TransactionState s = CurrentTransactionState;
ShowTransactionState("CleanupSubTransaction");
if (s->state != TRANS_ABORT)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "CleanupSubTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
AtSubCleanup_Portals(s->subTransactionId);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
CurrentResourceOwner = s->parent->curTransactionOwner;
CurTransactionResourceOwner = s->parent->curTransactionOwner;
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
if (s->curTransactionOwner)
ResourceOwnerDelete(s->curTransactionOwner);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
s->curTransactionOwner = NULL;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
AtSubCleanup_Memory();
s->state = TRANS_DEFAULT;
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
PopTransaction();
Further review of xact.c state machine for nested transactions. Fix problems with starting subtransactions inside already-failed transactions. Clean up some comments.
2004-07-01 22:11:03 +02:00
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* PushTransaction
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* Create transaction state stack entry for a subtransaction
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*
* The caller has to make sure to always reassign CurrentTransactionState
* if it has a local pointer to it after calling this function.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
static void
PushTransaction(void)
{
TransactionState p = CurrentTransactionState;
TransactionState s;
/*
* We keep subtransaction state nodes in TopTransactionContext.
*/
s = (TransactionState)
MemoryContextAllocZero(TopTransactionContext,
sizeof(TransactionStateData));
Standard pgindent run for 8.1.
2005-10-15 04:49:52 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* Assign a subtransaction ID, watching out for counter wraparound.
*/
currentSubTransactionId += 1;
if (currentSubTransactionId == InvalidSubTransactionId)
{
currentSubTransactionId -= 1;
pfree(s);
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot have more than 2^32-1 subtransactions in a transaction")));
}
Standard pgindent run for 8.1.
2005-10-15 04:49:52 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
/*
* We can now stack a minimally valid subtransaction without fear of
* failure.
*/
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
s->fullTransactionId = InvalidFullTransactionId; /* until assigned */
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->subTransactionId = currentSubTransactionId;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->parent = p;
s->nestingLevel = p->nestingLevel + 1;
Implement function-local GUC parameter settings, as per recent discussion. There are still some loose ends: I didn't do anything about the SET FROM CURRENT idea yet, and it's not real clear whether we are happy with the interaction of SET LOCAL with function-local settings. The documentation is a bit spartan, too.
2007-09-03 02:39:26 +02:00
s->gucNestLevel = NewGUCNestLevel();
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
s->savepointLevel = p->savepointLevel;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
s->state = TRANS_DEFAULT;
s->blockState = TBLOCK_SUBBEGIN;
Prevent indirect security attacks via changing session-local state within an allegedly immutable index function. It was previously recognized that we had to prevent such a function from executing SET/RESET ROLE/SESSION AUTHORIZATION, or it could trivially obtain the privileges of the session user. However, since there is in general no privilege checking for changes of session-local state, it is also possible for such a function to change settings in a way that might subvert later operations in the same session. Examples include changing search_path to cause an unexpected function to be called, or replacing an existing prepared statement with another one that will execute a function of the attacker's choosing. The present patch secures VACUUM, ANALYZE, and CREATE INDEX/REINDEX against these threats, which are the same places previously deemed to need protection against the SET ROLE issue. GUC changes are still allowed, since there are many useful cases for that, but we prevent security problems by forcing a rollback of any GUC change after completing the operation. Other cases are handled by throwing an error if any change is attempted; these include temp table creation, closing a cursor, and creating or deleting a prepared statement. (In 7.4, the infrastructure to roll back GUC changes doesn't exist, so we settle for rejecting changes of "search_path" in these contexts.) Original report and patch by Gurjeet Singh, additional analysis by Tom Lane. Security: CVE-2009-4136
2009-12-09 22:57:51 +01:00
GetUserIdAndSecContext(&s->prevUser, &s->prevSecContext);
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
s->prevXactReadOnly = XactReadOnly;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
s->parallelModeLevel = 0;
Replace XLOG_INCLUDE_XID flag with a more localized flag. Commit 0bead9af484c introduced XLOG_INCLUDE_XID flag to indicate that the WAL record contains subXID-to-topXID association. It uses that flag later to mark in CurrentTransactionState that top-xid is logged so that we should not try to log it again with the next WAL record in the current subtransaction. However, we can use a localized variable to pass that information. In passing, change the related function and variable names to make them consistent with what the code is actually doing. Author: Dilip Kumar Reviewed-by: Alvaro Herrera, Amit Kapila Discussion: https://postgr.es/m/E1mSoYz-0007Fh-D9@gemulon.postgresql.org
2021-11-02 04:05:29 +01:00
s->topXidLogged = false;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
CurrentTransactionState = s;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
* AbortSubTransaction and CleanupSubTransaction have to be able to cope
* with the subtransaction from here on out; in particular they should not
* assume that it necessarily has a transaction context, resource owner,
* or XID.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
}
/*
* PopTransaction
* Pop back to parent transaction state
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
*
* The caller has to make sure to always reassign CurrentTransactionState
* if it has a local pointer to it after calling this function.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
static void
PopTransaction(void)
{
TransactionState s = CurrentTransactionState;
if (s->state != TRANS_DEFAULT)
Fix subtransaction behavior for large objects, temp namespace, files, password/group files. Also allow read-only subtransactions of a read-write parent, but not vice versa. These are the reasonably noncontroversial parts of Alvaro's recent mop-up patch, plus further work on large objects to minimize use of the TopTransactionResourceOwner.
2004-07-28 16:23:31 +02:00
elog(WARNING, "PopTransaction while in %s state",
TransStateAsString(s->state));
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
if (s->parent == NULL)
elog(FATAL, "PopTransaction with no parent");
CurrentTransactionState = s->parent;
/* Let's just make sure CurTransactionContext is good */
CurTransactionContext = s->parent->curTransactionContext;
MemoryContextSwitchTo(CurTransactionContext);
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
/* Ditto for ResourceOwner links */
CurTransactionResourceOwner = s->parent->curTransactionOwner;
CurrentResourceOwner = s->parent->curTransactionOwner;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/* Free the old child structure */
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
if (s->name)
pfree(s->name);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
pfree(s);
}
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* EstimateTransactionStateSpace
* Estimate the amount of space that will be needed by
* SerializeTransactionState. It would be OK to overestimate slightly,
* but it's simple for us to work out the precise value, so we do.
*/
Size
EstimateTransactionStateSpace(void)
{
TransactionState s;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
Size nxids = 0;
Size size = SerializedTransactionStateHeaderSize;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
for (s = CurrentTransactionState; s != NULL; s = s->parent)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nxids = add_size(nxids, 1);
nxids = add_size(nxids, s->nChildXids);
}
Fix GetNewTransactionId()'s interaction with xidVacLimit. Commit ad308058 switched to returning a FullTransactionId, but failed to load the potentially updated value in the case where xidVacLimit is reached and we release and reacquire the lock. Repair, closing bug #15727. While reviewing that commit, also fix the size computation used by EstimateTransactionStateSize() and switch to the mul_size() macro traditionally used in such expressions. Author: Thomas Munro Reported-by: Roman Zharkov Discussion: https://postgr.es/m/15727-0be246e7d852d229%40postgresql.org
2019-04-12 04:53:38 +02:00
return add_size(size, mul_size(sizeof(TransactionId), nxids));
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
}
/*
* SerializeTransactionState
* Write out relevant details of our transaction state that will be
* needed by a parallel worker.
*
* We need to save and restore XactDeferrable, XactIsoLevel, and the XIDs
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
* associated with this transaction. These are serialized into a
* caller-supplied buffer big enough to hold the number of bytes reported by
* EstimateTransactionStateSpace(). We emit the XIDs in sorted order for the
* convenience of the receiving process.
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
*/
void
SerializeTransactionState(Size maxsize, char *start_address)
{
TransactionState s;
Size nxids = 0;
Size i = 0;
TransactionId *workspace;
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
SerializedTransactionState *result;
result = (SerializedTransactionState *) start_address;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
result->xactIsoLevel = XactIsoLevel;
result->xactDeferrable = XactDeferrable;
result->topFullTransactionId = XactTopFullTransactionId;
result->currentFullTransactionId =
CurrentTransactionState->fullTransactionId;
result->currentCommandId = currentCommandId;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/*
* If we're running in a parallel worker and launching a parallel worker
* of our own, we can just pass along the information that was passed to
* us.
*/
if (nParallelCurrentXids > 0)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
result->nParallelCurrentXids = nParallelCurrentXids;
memcpy(&result->parallelCurrentXids[0], ParallelCurrentXids,
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nParallelCurrentXids * sizeof(TransactionId));
return;
}
/*
* OK, we need to generate a sorted list of XIDs that our workers should
* view as current. First, figure out how many there are.
*/
for (s = CurrentTransactionState; s != NULL; s = s->parent)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
nxids = add_size(nxids, 1);
nxids = add_size(nxids, s->nChildXids);
}
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
Assert(SerializedTransactionStateHeaderSize + nxids * sizeof(TransactionId)
<= maxsize);
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
/* Copy them to our scratch space. */
workspace = palloc(nxids * sizeof(TransactionId));
for (s = CurrentTransactionState; s != NULL; s = s->parent)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
if (FullTransactionIdIsValid(s->fullTransactionId))
workspace[i++] = XidFromFullTransactionId(s->fullTransactionId);
Clean up assorted failures under clang's -fsanitize=undefined checks. Most of these are cases where we could call memcpy() or other libc functions with a NULL pointer and a zero count, which is forbidden by POSIX even though every production version of libc allows it. We've fixed such things before in a piecemeal way, but apparently never made an effort to try to get them all. I don't claim that this patch does so either, but it gets every failure I observe in check-world, using clang 12.0.1 on current RHEL8. numeric.c has a different issue that the sanitizer doesn't like: "ln(-1.0)" will compute log10(0) and then try to assign the resulting -Inf to an integer variable. We don't actually use the result in such a case, so there's no live bug. Back-patch to all supported branches, with the idea that we might start running a buildfarm member that tests this case. This includes back-patching c1132aae3 (Check the size in COPY_POINTER_FIELD), which previously silenced some of these issues in copyfuncs.c. Discussion: https://postgr.es/m/CALNJ-vT9r0DSsAOw9OXVJFxLENoVS_68kJ5x0p44atoYH+H4dg@mail.gmail.com
2022-03-04 00:13:24 +01:00
if (s->nChildXids > 0)
memcpy(&workspace[i], s->childXids,
s->nChildXids * sizeof(TransactionId));
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
i += s->nChildXids;
}
Assert(i == nxids);
/* Sort them. */
qsort(workspace, nxids, sizeof(TransactionId), xidComparator);
/* Copy data into output area. */
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
result->nParallelCurrentXids = nxids;
memcpy(&result->parallelCurrentXids[0], workspace,
nxids * sizeof(TransactionId));
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
}
/*
* StartParallelWorkerTransaction
* Start a parallel worker transaction, restoring the relevant
* transaction state serialized by SerializeTransactionState.
*/
void
StartParallelWorkerTransaction(char *tstatespace)
{
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
SerializedTransactionState *tstate;
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
Assert(CurrentTransactionState->blockState == TBLOCK_DEFAULT);
StartTransaction();
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
tstate = (SerializedTransactionState *) tstatespace;
XactIsoLevel = tstate->xactIsoLevel;
XactDeferrable = tstate->xactDeferrable;
XactTopFullTransactionId = tstate->topFullTransactionId;
CurrentTransactionState->fullTransactionId =
tstate->currentFullTransactionId;
currentCommandId = tstate->currentCommandId;
nParallelCurrentXids = tstate->nParallelCurrentXids;
ParallelCurrentXids = &tstate->parallelCurrentXids[0];
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
CurrentTransactionState->blockState = TBLOCK_PARALLEL_INPROGRESS;
}
/*
* EndParallelWorkerTransaction
* End a parallel worker transaction.
*/
void
EndParallelWorkerTransaction(void)
{
Assert(CurrentTransactionState->blockState == TBLOCK_PARALLEL_INPROGRESS);
CommitTransaction();
CurrentTransactionState->blockState = TBLOCK_DEFAULT;
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
/*
* ShowTransactionState
* Debug support
*/
static void
ShowTransactionState(const char *str)
{
/* skip work if message will definitely not be printed */
Centralize logic for skipping useless ereport/elog calls. While ereport() and elog() themselves are quite cheap when the error message level is too low to be printed, some places need to do substantial work before they can call those macros at all. To allow optimizing away such setup work when nothing is to be printed, make elog.c export a new function message_level_is_interesting(elevel) that reports whether ereport/elog will do anything. Make use of that in various places that had ad-hoc direct tests of log_min_messages etc. Also teach ProcSleep to use it to avoid some work. (There may well be other places that could usefully use this; I didn't search hard.) Within elog.c, refactor a little bit to avoid having duplicate copies of the policy-setting logic. When that code was written, we weren't relying on the availability of inline functions; so it had some duplications in the name of efficiency, which I got rid of. Alvaro Herrera and Tom Lane Discussion: https://postgr.es/m/129515.1606166429@sss.pgh.pa.us
2020-11-24 01:04:07 +01:00
if (message_level_is_interesting(DEBUG5))
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
ShowTransactionStateRec(str, CurrentTransactionState);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
/*
* ShowTransactionStateRec
* Recursive subroutine for ShowTransactionState
*/
static void
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
ShowTransactionStateRec(const char *str, TransactionState s)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
{
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
StringInfoData buf;
initStringInfo(&buf);
if (s->nChildXids > 0)
{
int i;
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
appendStringInfo(&buf, ", children: %u", s->childXids[0]);
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
for (i = 1; i < s->nChildXids; i++)
appendStringInfo(&buf, " %u", s->childXids[i]);
}
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
if (s->parent)
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
ShowTransactionStateRec(str, s->parent);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
ereport(DEBUG5,
(errmsg_internal("%s(%d) name: %s; blockState: %s; state: %s, xid/subid/cid: %u/%u/%u%s%s",
str, s->nestingLevel,
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
PointerIsValid(s->name) ? s->name : "unnamed",
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
BlockStateAsString(s->blockState),
TransStateAsString(s->state),
Use FullTransactionId for the transaction stack. Provide GetTopFullTransactionId() and GetCurrentFullTransactionId(). The intended users of these interfaces are access methods that use xids for visibility checks but don't want to have to go back and "freeze" existing references some time later before the 32 bit xid counter wraps around. Use a new struct to serialize the transaction state for parallel query, because FullTransactionId doesn't fit into the previous serialization scheme very well. Author: Thomas Munro Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:59:19 +01:00
(unsigned int) XidFromFullTransactionId(s->fullTransactionId),
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
(unsigned int) s->subTransactionId,
(unsigned int) currentCommandId,
Avoid incrementing the CommandCounter when CommandCounterIncrement is called but no database changes have been made since the last CommandCounterIncrement. This should result in a significant improvement in the number of "commands" that can typically be performed within a transaction before hitting the 2^32 CommandId size limit. In particular this buys back (and more) the possible adverse consequences of my previous patch to fix plan caching behavior. The implementation requires tracking whether the current CommandCounter value has been "used" to mark any tuples. CommandCounter values stored into snapshots are presumed not to be used for this purpose. This requires some small executor changes, since the executor used to conflate the curcid of the snapshot it was using with the command ID to mark output tuples with. Separating these concepts allows some small simplifications in executor APIs. Something for the TODO list: look into having CommandCounterIncrement not do AcceptInvalidationMessages. It seems fairly bogus to be doing it there, but exactly where to do it instead isn't clear, and I'm disinclined to mess with asynchronous behavior during late beta.
2007-11-30 22:22:54 +01:00
currentCommandIdUsed ? " (used)" : "",
Remove or reduce verbosity of some debug messages. The debug messages that merely print StartTransactionCommand, CommitTransactionCommand, ProcessUtilty, or ProcessQuery with no additional details seem to be useless. Get rid of them. The transaction status messages produced by ShowTransactionState are occasionally useful, but they are extremely verbose, producing multiple lines of log output every time they fire, which can happens multiple times per transaction. So, reduce the level to DEBUG5; avoid emitting an extra line just to explain which debug point is at issue; and tighten up the rest of the message so it doesn't use quite so much horizontal space. With these changes, it's possible to run a somewhat busy system with a log level even as high as DEBUG4, whereas previously anything above DEBUG2 would flood the log with output that probably wasn't really all that useful.
2016-11-17 22:54:24 +01:00
buf.data)));
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
pfree(buf.data);
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
/*
* BlockStateAsString
* Debug support
*/
static const char *
BlockStateAsString(TBlockState blockState)
{
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (blockState)
{
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_DEFAULT:
return "DEFAULT";
case TBLOCK_STARTED:
return "STARTED";
case TBLOCK_BEGIN:
return "BEGIN";
case TBLOCK_INPROGRESS:
return "INPROGRESS";
Fix handling of savepoint commands within multi-statement Query strings. Issuing a savepoint-related command in a Query message that contains multiple SQL statements led to a FATAL exit with a complaint about "unexpected state STARTED". This is a shortcoming of commit 4f896dac1, which attempted to prevent such misbehaviors in multi-statement strings; its quick hack of marking the individual statements as "not top-level" does the wrong thing in this case, and isn't a very accurate description of the situation anyway. To fix, let's introduce into xact.c an explicit model of what happens for multi-statement Query strings. This is an "implicit transaction block in progress" state, which for many purposes works like the normal TBLOCK_INPROGRESS state --- in particular, IsTransactionBlock returns true, causing the desired result that PreventTransactionChain will throw error. But in case of error abort it works like TBLOCK_STARTED, allowing the transaction to be cancelled without need for an explicit ROLLBACK command. Commit 4f896dac1 is reverted in toto, so that we go back to treating the individual statements as "top level". We could have left it as-is, but this allows sharpening the error message for PreventTransactionChain calls inside functions. Except for getting a normal error instead of a FATAL exit for savepoint commands, this patch should result in no user-visible behavioral change (other than that one error message rewording). There are some things we might want to do in the line of changing the appearance or wording of error and warning messages around this behavior, which would be much simpler to do now that it's an explicitly modeled state. But I haven't done them here. Although this fixes a long-standing bug, no backpatch. The consequences of the bug don't seem severe enough to justify the risk that this commit itself creates some new issue. Patch by me, but it owes something to previous investigation by Takayuki Tsunakawa, who also reported the bug in the first place. Also thanks to Michael Paquier for reviewing. Discussion: https://postgr.es/m/0A3221C70F24FB45833433255569204D1F6BE40D@G01JPEXMBYT05
2017-09-07 15:49:55 +02:00
case TBLOCK_IMPLICIT_INPROGRESS:
return "IMPLICIT_INPROGRESS";
Create an infrastructure for parallel computation in PostgreSQL. This does four basic things. First, it provides convenience routines to coordinate the startup and shutdown of parallel workers. Second, it synchronizes various pieces of state (e.g. GUCs, combo CID mappings, transaction snapshot) from the parallel group leader to the worker processes. Third, it prohibits various operations that would result in unsafe changes to that state while parallelism is active. Finally, it propagates events that would result in an ErrorResponse, NoticeResponse, or NotifyResponse message being sent to the client from the parallel workers back to the master, from which they can then be sent on to the client. Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke. Suggestions and review from Andres Freund, Heikki Linnakangas, Noah Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
2015-04-30 21:02:14 +02:00
case TBLOCK_PARALLEL_INPROGRESS:
return "PARALLEL_INPROGRESS";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_END:
return "END";
case TBLOCK_ABORT:
return "ABORT";
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_END:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "ABORT_END";
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_ABORT_PENDING:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "ABORT_PENDING";
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TBLOCK_PREPARE:
return "PREPARE";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBBEGIN:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBBEGIN";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBINPROGRESS:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBINPROGRESS";
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBRELEASE:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBRELEASE";
Remove O(N^2) performance issue with multiple SAVEPOINTs. Subtransaction locks now released en masse at main commit, rather than repeatedly re-scanning for locks as we ascend the nested transaction tree. Split transaction state TBLOCK_SUBEND into two states, TBLOCK_SUBCOMMIT and TBLOCK_SUBRELEASE to allow the commit path to be optimised using the existing code in ResourceOwnerRelease() which appears to have been intended for this usage, judging from comments therein.
2011-07-19 18:21:24 +02:00
case TBLOCK_SUBCOMMIT:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBCOMMIT";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TBLOCK_SUBABORT:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBABORT";
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_END:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBABORT_END";
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
case TBLOCK_SUBABORT_PENDING:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBABORT_PENDING";
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBRESTART:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBRESTART";
Restructure subtransaction handling to reduce resource consumption, as per recent discussions. Invent SubTransactionIds that are managed like CommandIds (ie, counter is reset at start of each top transaction), and use these instead of TransactionIds to keep track of subtransaction status in those modules that need it. This means that a subtransaction does not need an XID unless it actually inserts/modifies rows in the database. Accordingly, don't assign it an XID nor take a lock on the XID until it tries to do that. This saves a lot of overhead for subtransactions that are only used for error recovery (eg plpgsql exceptions). Also, arrange to release a subtransaction's XID lock as soon as the subtransaction exits, in both the commit and abort cases. This avoids holding many unique locks after a long series of subtransactions. The price is some additional overhead in XactLockTableWait, but that seems acceptable. Finally, restructure the state machine in xact.c to have a more orthogonal set of states for subtransactions.
2004-09-16 18:58:44 +02:00
case TBLOCK_SUBABORT_RESTART:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "SUBABORT_RESTART";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
return "UNRECOGNIZED";
}
/*
* TransStateAsString
* Debug support
*/
static const char *
TransStateAsString(TransState state)
{
Replace nested-BEGIN syntax for subtransactions with spec-compliant SAVEPOINT/RELEASE/ROLLBACK-TO syntax. (Alvaro) Cause COMMIT of a failed transaction to report ROLLBACK instead of COMMIT in its command tag. (Tom) Fix a few loose ends in the nested-transactions stuff.
2004-07-27 07:11:48 +02:00
switch (state)
{
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TRANS_DEFAULT:
return "DEFAULT";
case TRANS_START:
return "START";
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TRANS_INPROGRESS:
Change transaction state debug strings to match enum symbols In some cases, these were different for no apparent reason, making debugging unnecessarily mysterious. Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-02-17 16:19:21 +01:00
return "INPROGRESS";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
case TRANS_COMMIT:
return "COMMIT";
case TRANS_ABORT:
return "ABORT";
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
case TRANS_PREPARE:
return "PREPARE";
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
return "UNRECOGNIZED";
}
/*
* xactGetCommittedChildren
*
* Gets the list of committed children of the current transaction. The return
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
* value is the number of child transactions. *ptr is set to point to an
* array of TransactionIds. The array is allocated in TopTransactionContext;
* the caller should *not* pfree() it (this is a change from pre-8.4 code!).
* If there are no subxacts, *ptr is set to NULL.
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*/
int
Invent ResourceOwner mechanism as per my recent proposal, and use it to keep track of portal-related resources separately from transaction-related resources. This allows cursors to work in a somewhat sane fashion with nested transactions. For now, cursor behavior is non-subtransactional, that is a cursor's state does not roll back if you abort a subtransaction that fetched from the cursor. We might want to change that later.
2004-07-17 05:32:14 +02:00
xactGetCommittedChildren(TransactionId **ptr)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
{
TransactionState s = CurrentTransactionState;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
if (s->nChildXids == 0)
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
*ptr = NULL;
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
else
*ptr = s->childXids;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
Fix TransactionIdIsCurrentTransactionId() to use binary search instead of linear search when checking child-transaction XIDs. This makes for an important speedup in transactions that have large numbers of children, as in a recent example from Craig Ringer. We can also get rid of an ugly kluge that represented lists of TransactionIds as lists of OIDs. Heikki Linnakangas
2008-03-17 03:18:55 +01:00
return s->nChildXids;
Nested transactions. There is still much left to do, especially on the performance front, but with feature freeze upon us I think it's time to drive a stake in the ground and say that this will be in 7.5. Alvaro Herrera, with some help from Tom Lane.
2004-07-01 02:52:04 +02:00
}
Update xact.c comments for clarity.
2002-11-13 04:12:05 +01:00
/*
* XLOG support routines
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
/*
* Log the commit record for a plain or twophase transaction commit.
*
* A 2pc commit will be emitted when twophase_xid is valid, a plain one
* otherwise.
*/
XLogRecPtr
XactLogCommitRecord(TimestampTz commit_time,
int nsubxacts, TransactionId *subxacts,
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
int nrels, RelFileLocator *rels,
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
int ndroppedstats, xl_xact_stats_item *droppedstats,
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
int nmsgs, SharedInvalidationMessage *msgs,
Remove dead forceSync parameter of XactLogCommitRecord(). The function has been reading global variable forceSyncCommit, mirroring the intent of the caller that passed forceSync=forceSyncCommit. The other caller, RecordTransactionCommitPrepared(), passed false. Since COMMIT PREPARED can't share a transaction with any command, it certainly doesn't share a transaction with a command that sets forceSyncCommit. Reviewed by Michael Paquier. Discussion: https://postgr.es/m/20200617032615.GC2916904@rfd.leadboat.com
2020-06-20 10:25:40 +02:00
bool relcacheInval,
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
int xactflags, TransactionId twophase_xid,
const char *twophase_gid)
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
{
xl_xact_commit xlrec;
xl_xact_xinfo xl_xinfo;
xl_xact_dbinfo xl_dbinfo;
xl_xact_subxacts xl_subxacts;
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
xl_xact_relfilelocators xl_relfilelocators;
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
xl_xact_stats_items xl_dropped_stats;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xl_xact_invals xl_invals;
xl_xact_twophase xl_twophase;
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
xl_xact_origin xl_origin;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
uint8 info;
Assert(CritSectionCount > 0);
xl_xinfo.xinfo = 0;
/* decide between a plain and 2pc commit */
if (!TransactionIdIsValid(twophase_xid))
info = XLOG_XACT_COMMIT;
else
info = XLOG_XACT_COMMIT_PREPARED;
/* First figure out and collect all the information needed */
xlrec.xact_time = commit_time;
if (relcacheInval)
xl_xinfo.xinfo |= XACT_COMPLETION_UPDATE_RELCACHE_FILE;
if (forceSyncCommit)
xl_xinfo.xinfo |= XACT_COMPLETION_FORCE_SYNC_COMMIT;
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
if ((xactflags & XACT_FLAGS_ACQUIREDACCESSEXCLUSIVELOCK))
xl_xinfo.xinfo |= XACT_XINFO_HAS_AE_LOCKS;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
Add new replication mode synchronous_commit = 'remote_apply'. In this mode, the master waits for the transaction to be applied on the remote side, not just written to disk. That means that you can count on a transaction started on the standby to see all commits previously acknowledged by the master. To make this work, the standby sends a reply after replaying each commit record generated with synchronous_commit >= 'remote_apply'. This introduces a small inefficiency: the extra replies will be sent even by standbys that aren't the current synchronous standby. But previously-existing synchronous_commit levels make no attempt at all to optimize which replies are sent based on what the primary cares about, so this is no worse, and at least avoids any extra replies for people not using the feature at all. Thomas Munro, reviewed by Michael Paquier and by me. Some additional tweaks by me.
2016-03-30 03:16:12 +02:00
/*
* Check if the caller would like to ask standbys for immediate feedback
* once this commit is applied.
*/
if (synchronous_commit >= SYNCHRONOUS_COMMIT_REMOTE_APPLY)
xl_xinfo.xinfo |= XACT_COMPLETION_APPLY_FEEDBACK;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
/*
* Relcache invalidations requires information about the current database
* and so does logical decoding.
*/
if (nmsgs > 0 || XLogLogicalInfoActive())
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_DBINFO;
xl_dbinfo.dbId = MyDatabaseId;
xl_dbinfo.tsId = MyDatabaseTableSpace;
}
if (nsubxacts > 0)
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_SUBXACTS;
xl_subxacts.nsubxacts = nsubxacts;
}
if (nrels > 0)
{
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
xl_xinfo.xinfo |= XACT_XINFO_HAS_RELFILELOCATORS;
xl_relfilelocators.nrels = nrels;
Mark commit and abort WAL records with XLR_SPECIAL_REL_UPDATE. If a commit or abort record includes "dropped relfilenodes", then replaying the record will remove data files. That is surely a "special rel update", but the records were not marked as such. Fix that, teach pg_rewind to expect and ignore them, and add a test case to cover it. It's always been like this, but no backporting for fear of breaking existing applications. If an application parsed the WAL but was not handling commit/abort records, it would stop working. That might be a good thing if it really needed to handle the dropped rels, but it will be caught when the application is updated to work with PostgreSQL v14 anyway. Discussion: https://www.postgresql.org/message-id/07b33e2c-46a6-86a1-5f9e-a7da73fddb95%40iki.fi Reviewed-by: Amit Kapila, Michael Paquier
2020-08-17 09:52:58 +02:00
info |= XLR_SPECIAL_REL_UPDATE;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (ndroppedstats > 0)
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_DROPPED_STATS;
xl_dropped_stats.nitems = ndroppedstats;
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (nmsgs > 0)
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_INVALS;
xl_invals.nmsgs = nmsgs;
}
if (TransactionIdIsValid(twophase_xid))
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_TWOPHASE;
xl_twophase.xid = twophase_xid;
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
Assert(twophase_gid != NULL);
if (XLogLogicalInfoActive())
xl_xinfo.xinfo |= XACT_XINFO_HAS_GID;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
/* dump transaction origin information */
Fix "sesssion" typo It was introduced alongside replication origins, by commit 5aa2350426c, so backpatch to 9.5. Pointed out by Fujii Masao
2015-09-29 00:13:42 +02:00
if (replorigin_session_origin != InvalidRepOriginId)
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_ORIGIN;
Fix "sesssion" typo It was introduced alongside replication origins, by commit 5aa2350426c, so backpatch to 9.5. Pointed out by Fujii Masao
2015-09-29 00:13:42 +02:00
xl_origin.origin_lsn = replorigin_session_origin_lsn;
xl_origin.origin_timestamp = replorigin_session_origin_timestamp;
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (xl_xinfo.xinfo != 0)
info |= XLOG_XACT_HAS_INFO;
/* Then include all the collected data into the commit record. */
XLogBeginInsert();
XLogRegisterData((char *) (&xlrec), sizeof(xl_xact_commit));
if (xl_xinfo.xinfo != 0)
XLogRegisterData((char *) (&xl_xinfo.xinfo), sizeof(xl_xinfo.xinfo));
if (xl_xinfo.xinfo & XACT_XINFO_HAS_DBINFO)
XLogRegisterData((char *) (&xl_dbinfo), sizeof(xl_dbinfo));
if (xl_xinfo.xinfo & XACT_XINFO_HAS_SUBXACTS)
{
XLogRegisterData((char *) (&xl_subxacts),
MinSizeOfXactSubxacts);
XLogRegisterData((char *) subxacts,
nsubxacts * sizeof(TransactionId));
}
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_RELFILELOCATORS)
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
{
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
XLogRegisterData((char *) (&xl_relfilelocators),
MinSizeOfXactRelfileLocators);
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
XLogRegisterData((char *) rels,
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
nrels * sizeof(RelFileLocator));
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_DROPPED_STATS)
{
XLogRegisterData((char *) (&xl_dropped_stats),
MinSizeOfXactStatsItems);
XLogRegisterData((char *) droppedstats,
ndroppedstats * sizeof(xl_xact_stats_item));
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_INVALS)
{
XLogRegisterData((char *) (&xl_invals), MinSizeOfXactInvals);
XLogRegisterData((char *) msgs,
nmsgs * sizeof(SharedInvalidationMessage));
}
if (xl_xinfo.xinfo & XACT_XINFO_HAS_TWOPHASE)
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
{
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
XLogRegisterData((char *) (&xl_twophase), sizeof(xl_xact_twophase));
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_GID)
More unconstify use Replace casts whose only purpose is to cast away const with the unconstify() macro. Discussion: https://www.postgresql.org/message-id/flat/53a28052-f9f3-1808-fed9-460fd43035ab%402ndquadrant.com
2019-01-29 01:16:24 +01:00
XLogRegisterData(unconstify(char *, twophase_gid), strlen(twophase_gid) + 1);
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_ORIGIN)
XLogRegisterData((char *) (&xl_origin), sizeof(xl_xact_origin));
/* we allow filtering by xacts */
Skip checkpoints, archiving on idle systems. Some background activity (like checkpoints, archive timeout, standby snapshots) is not supposed to happen on an idle system. Unfortunately so far it was not easy to determine when a system is idle, which defeated some of the attempts to avoid redundant activity on an idle system. To make that easier, allow to make individual WAL insertions as not being "important". By checking whether any important activity happened since the last time an activity was performed, it now is easy to check whether some action needs to be repeated. Use the new facility for checkpoints, archive timeout and standby snapshots. The lack of a facility causes some issues in older releases, but in my opinion the consequences (superflous checkpoints / archived segments) aren't grave enough to warrant backpatching. Author: Michael Paquier, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Amit Kapila, Kyotaro HORIGUCHI Bug: #13685 Discussion: https://www.postgresql.org/message-id/20151016203031.3019.72930@wrigleys.postgresql.org https://www.postgresql.org/message-id/CAB7nPqQcPqxEM3S735Bd2RzApNqSNJVietAC=6kfkYv_45dKwA@mail.gmail.com Backpatch: -
2016-12-22 20:31:50 +01:00
XLogSetRecordFlags(XLOG_INCLUDE_ORIGIN);
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
return XLogInsert(RM_XACT_ID, info);
}
/*
* Log the commit record for a plain or twophase transaction abort.
*
* A 2pc abort will be emitted when twophase_xid is valid, a plain one
* otherwise.
*/
XLogRecPtr
XactLogAbortRecord(TimestampTz abort_time,
int nsubxacts, TransactionId *subxacts,
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
int nrels, RelFileLocator *rels,
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
int ndroppedstats, xl_xact_stats_item *droppedstats,
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
int xactflags, TransactionId twophase_xid,
const char *twophase_gid)
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
{
xl_xact_abort xlrec;
xl_xact_xinfo xl_xinfo;
xl_xact_subxacts xl_subxacts;
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
xl_xact_relfilelocators xl_relfilelocators;
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
xl_xact_stats_items xl_dropped_stats;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xl_xact_twophase xl_twophase;
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
xl_xact_dbinfo xl_dbinfo;
xl_xact_origin xl_origin;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
uint8 info;
Assert(CritSectionCount > 0);
xl_xinfo.xinfo = 0;
/* decide between a plain and 2pc abort */
if (!TransactionIdIsValid(twophase_xid))
info = XLOG_XACT_ABORT;
else
info = XLOG_XACT_ABORT_PREPARED;
/* First figure out and collect all the information needed */
xlrec.xact_time = abort_time;
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
if ((xactflags & XACT_FLAGS_ACQUIREDACCESSEXCLUSIVELOCK))
xl_xinfo.xinfo |= XACT_XINFO_HAS_AE_LOCKS;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (nsubxacts > 0)
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_SUBXACTS;
xl_subxacts.nsubxacts = nsubxacts;
}
if (nrels > 0)
{
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
xl_xinfo.xinfo |= XACT_XINFO_HAS_RELFILELOCATORS;
xl_relfilelocators.nrels = nrels;
Mark commit and abort WAL records with XLR_SPECIAL_REL_UPDATE. If a commit or abort record includes "dropped relfilenodes", then replaying the record will remove data files. That is surely a "special rel update", but the records were not marked as such. Fix that, teach pg_rewind to expect and ignore them, and add a test case to cover it. It's always been like this, but no backporting for fear of breaking existing applications. If an application parsed the WAL but was not handling commit/abort records, it would stop working. That might be a good thing if it really needed to handle the dropped rels, but it will be caught when the application is updated to work with PostgreSQL v14 anyway. Discussion: https://www.postgresql.org/message-id/07b33e2c-46a6-86a1-5f9e-a7da73fddb95%40iki.fi Reviewed-by: Amit Kapila, Michael Paquier
2020-08-17 09:52:58 +02:00
info |= XLR_SPECIAL_REL_UPDATE;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (ndroppedstats > 0)
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_DROPPED_STATS;
xl_dropped_stats.nitems = ndroppedstats;
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (TransactionIdIsValid(twophase_xid))
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_TWOPHASE;
xl_twophase.xid = twophase_xid;
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
Assert(twophase_gid != NULL);
if (XLogLogicalInfoActive())
xl_xinfo.xinfo |= XACT_XINFO_HAS_GID;
}
if (TransactionIdIsValid(twophase_xid) && XLogLogicalInfoActive())
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_DBINFO;
xl_dbinfo.dbId = MyDatabaseId;
xl_dbinfo.tsId = MyDatabaseTableSpace;
}
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
/*
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
* Dump transaction origin information. We need this during recovery to
* update the replication origin progress.
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
*/
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
if (replorigin_session_origin != InvalidRepOriginId)
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
{
xl_xinfo.xinfo |= XACT_XINFO_HAS_ORIGIN;
xl_origin.origin_lsn = replorigin_session_origin_lsn;
xl_origin.origin_timestamp = replorigin_session_origin_timestamp;
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
if (xl_xinfo.xinfo != 0)
info |= XLOG_XACT_HAS_INFO;
/* Then include all the collected data into the abort record. */
XLogBeginInsert();
XLogRegisterData((char *) (&xlrec), MinSizeOfXactAbort);
if (xl_xinfo.xinfo != 0)
XLogRegisterData((char *) (&xl_xinfo), sizeof(xl_xinfo));
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_DBINFO)
XLogRegisterData((char *) (&xl_dbinfo), sizeof(xl_dbinfo));
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_SUBXACTS)
{
XLogRegisterData((char *) (&xl_subxacts),
MinSizeOfXactSubxacts);
XLogRegisterData((char *) subxacts,
nsubxacts * sizeof(TransactionId));
}
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_RELFILELOCATORS)
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
{
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
XLogRegisterData((char *) (&xl_relfilelocators),
MinSizeOfXactRelfileLocators);
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
XLogRegisterData((char *) rels,
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
nrels * sizeof(RelFileLocator));
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
}
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_DROPPED_STATS)
{
XLogRegisterData((char *) (&xl_dropped_stats),
MinSizeOfXactStatsItems);
XLogRegisterData((char *) droppedstats,
ndroppedstats * sizeof(xl_xact_stats_item));
}
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_TWOPHASE)
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
{
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
XLogRegisterData((char *) (&xl_twophase), sizeof(xl_xact_twophase));
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
if (xl_xinfo.xinfo & XACT_XINFO_HAS_GID)
More unconstify use Replace casts whose only purpose is to cast away const with the unconstify() macro. Discussion: https://www.postgresql.org/message-id/flat/53a28052-f9f3-1808-fed9-460fd43035ab%402ndquadrant.com
2019-01-29 01:16:24 +01:00
XLogRegisterData(unconstify(char *, twophase_gid), strlen(twophase_gid) + 1);
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
}
if (xl_xinfo.xinfo & XACT_XINFO_HAS_ORIGIN)
XLogRegisterData((char *) (&xl_origin), sizeof(xl_xact_origin));
Perform apply of large transactions by parallel workers. Currently, for large transactions, the publisher sends the data in multiple streams (changes divided into chunks depending upon logical_decoding_work_mem), and then on the subscriber-side, the apply worker writes the changes into temporary files and once it receives the commit, it reads from those files and applies the entire transaction. To improve the performance of such transactions, we can instead allow them to be applied via parallel workers. In this approach, we assign a new parallel apply worker (if available) as soon as the xact's first stream is received and the leader apply worker will send changes to this new worker via shared memory. The parallel apply worker will directly apply the change instead of writing it to temporary files. However, if the leader apply worker times out while attempting to send a message to the parallel apply worker, it will switch to "partial serialize" mode - in this mode, the leader serializes all remaining changes to a file and notifies the parallel apply workers to read and apply them at the end of the transaction. We use a non-blocking way to send the messages from the leader apply worker to the parallel apply to avoid deadlocks. We keep this parallel apply assigned till the transaction commit is received and also wait for the worker to finish at commit. This preserves commit ordering and avoid writing to and reading from files in most cases. We still need to spill if there is no worker available. This patch also extends the SUBSCRIPTION 'streaming' parameter so that the user can control whether to apply the streaming transaction in a parallel apply worker or spill the change to disk. The user can set the streaming parameter to 'on/off', or 'parallel'. The parameter value 'parallel' means the streaming will be applied via a parallel apply worker, if available. The parameter value 'on' means the streaming transaction will be spilled to disk. The default value is 'off' (same as current behaviour). In addition, the patch extends the logical replication STREAM_ABORT message so that abort_lsn and abort_time can also be sent which can be used to update the replication origin in parallel apply worker when the streaming transaction is aborted. Because this message extension is needed to support parallel streaming, parallel streaming is not supported for publications on servers < PG16. Author: Hou Zhijie, Wang wei, Amit Kapila with design inputs from Sawada Masahiko Reviewed-by: Sawada Masahiko, Peter Smith, Dilip Kumar, Shi yu, Kuroda Hayato, Shveta Mallik Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-09 02:30:39 +01:00
/* Include the replication origin */
XLogSetRecordFlags(XLOG_INCLUDE_ORIGIN);
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
return XLogInsert(RM_XACT_ID, info);
}
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
Stamp HEAD as 9.0devel, and update various places that were referring to 8.5 (hope I got 'em all). Per discussion, this release will be 9.0 not 8.5.
2010-02-17 05:19:41 +01:00
* Before 9.0 this was a fairly short function, but now it performs many
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
* actions for which the order of execution is critical.
*/
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
static void
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xact_redo_commit(xl_xact_parsed_commit *parsed,
TransactionId xid,
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
XLogRecPtr lsn,
RepOriginId origin_id)
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
{
TransactionId max_xid;
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
TimestampTz commit_time;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
Assert(TransactionIdIsValid(xid));
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
max_xid = TransactionIdLatest(xid, parsed->nsubxacts, parsed->subxacts);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
Rename VariableCacheData.nextFullXid to nextXid. Including Full in variable names duplicates the type information and leads to overly long names. As FullTransactionId cannot accidentally be casted to TransactionId that does not seem necessary. Author: Andres Freund Discussion: https://postgr.es/m/20200724011143.jccsyvsvymuiqfxu@alap3.anarazel.de
2020-08-11 20:25:23 +02:00
/* Make sure nextXid is beyond any XID mentioned in the record. */
Add basic infrastructure for 64 bit transaction IDs. Instead of inferring epoch progress from xids and checkpoints, introduce a 64 bit FullTransactionId type and use it to track xid generation. This fixes an unlikely bug where the epoch is reported incorrectly if the range of active xids wraps around more than once between checkpoints. The only user-visible effect of this commit is to correct the epoch used by txid_current() and txid_status(), also visible with pg_controldata, in those rare circumstances. It also creates some basic infrastructure so that later patches can use 64 bit transaction IDs in more places. The new type is a struct that we pass by value, as a form of strong typedef. This prevents the sort of accidental confusion between TransactionId and FullTransactionId that would be possible if we were to use a plain old uint64. Author: Thomas Munro Reported-by: Amit Kapila Reviewed-by: Andres Freund, Tom Lane, Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:34:43 +01:00
AdvanceNextFullTransactionIdPastXid(max_xid);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
Don't use !! but != 0/NULL to force boolean evaluation. I introduced several uses of !! to force bit arithmetic to be boolean, but per discussion the project prefers != 0/NULL. Discussion: CA+TgmoZP5KakLGP6B4vUjgMBUW0woq_dJYi0paOz-My0Hwt_vQ@mail.gmail.com
2016-03-27 17:17:00 +02:00
Assert(((parsed->xinfo & XACT_XINFO_HAS_ORIGIN) == 0) ==
(origin_id == InvalidRepOriginId));
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
if (parsed->xinfo & XACT_XINFO_HAS_ORIGIN)
commit_time = parsed->origin_timestamp;
else
commit_time = parsed->xact_time;
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
/* Set the transaction commit timestamp and metadata */
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionTreeSetCommitTsData(xid, parsed->nsubxacts, parsed->subxacts,
Remove COMMIT_TS_SETTS record. Commit 438fc4a39c prevented the WAL replay from writing COMMIT_TS_SETTS record. By this change there is no code that generates COMMIT_TS_SETTS record in PostgreSQL core. Also we can think that there are no extensions using the record because we've not received so far any complaints about the issue that commit 438fc4a39c fixed. Therefore this commit removes COMMIT_TS_SETTS record and its related code. Even without this record, the timestamp required for commit timestamp feature can be acquired from the COMMIT record. Bump WAL page magic. Reported-by: lx zou <zoulx1982@163.com> Author: Fujii Masao Reviewed-by: Alvaro Herrera Discussion: https://postgr.es/m/16931-620d0f2fdc6108f1@postgresql.org
2021-04-11 17:00:18 +02:00
commit_time, origin_id);
Keep track of transaction commit timestamps Transactions can now set their commit timestamp directly as they commit, or an external transaction commit timestamp can be fed from an outside system using the new function TransactionTreeSetCommitTsData(). This data is crash-safe, and truncated at Xid freeze point, same as pg_clog. This module is disabled by default because it causes a performance hit, but can be enabled in postgresql.conf requiring only a server restart. A new test in src/test/modules is included. Catalog version bumped due to the new subdirectory within PGDATA and a couple of new SQL functions. Authors: Álvaro Herrera and Petr Jelínek Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven Singer, Peter Eisentraut
2014-12-03 15:53:02 +01:00
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
if (standbyState == STANDBY_DISABLED)
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
{
/*
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
* Mark the transaction committed in pg_xact.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionIdCommitTree(xid, parsed->nsubxacts, parsed->subxacts);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
}
else
{
/*
Remove old-style VACUUM FULL (which was known for a little while as VACUUM FULL INPLACE), along with a boatload of subsidiary code and complexity. Per discussion, the use case for this method of vacuuming is no longer large enough to justify maintaining it; not to mention that we don't wish to invest the work that would be needed to make it play nicely with Hot Standby. Aside from the code directly related to old-style VACUUM FULL, this commit removes support for certain WAL record types that could only be generated within VACUUM FULL, redirect-pointer removal in heap_page_prune, and nontransactional generation of cache invalidation sinval messages (the last being the sticking point for Hot Standby). We still have to retain all code that copes with finding HEAP_MOVED_OFF and HEAP_MOVED_IN flag bits on existing tuples. This can't be removed as long as we want to support in-place update from pre-9.0 databases.
2010-02-08 05:33:55 +01:00
* If a transaction completion record arrives that has as-yet
* unobserved subtransactions then this will not have been fully
* handled by the call to RecordKnownAssignedTransactionIds() in the
* main recovery loop in xlog.c. So we need to do bookkeeping again to
* cover that case. This is confusing and it is easy to think this
* call is irrelevant, which has happened three times in development
* already. Leave it in.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
RecordKnownAssignedTransactionIds(max_xid);
/*
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
* Mark the transaction committed in pg_xact. We use async commit
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
* protocol during recovery to provide information on database
* consistency for when users try to set hint bits. It is important
* that we do not set hint bits until the minRecoveryPoint is past
* this commit record. This ensures that if we crash we don't see hint
* bits set on changes made by transactions that haven't yet
* recovered. It's unlikely but it's good to be safe.
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionIdAsyncCommitTree(xid, parsed->nsubxacts, parsed->subxacts, lsn);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* We must mark clog before we update the ProcArray.
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
ExpireTreeKnownAssignedTransactionIds(xid, parsed->nsubxacts, parsed->subxacts, max_xid);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* Send any cache invalidations attached to the commit. We must
* maintain the same order of invalidation then release locks as
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
* occurs in CommitTransaction().
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
ProcessCommittedInvalidationMessages(parsed->msgs, parsed->nmsgs,
XactCompletionRelcacheInitFileInval(parsed->xinfo),
parsed->dbId, parsed->tsId);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* Release locks, if any. We do this for both two phase and normal one
* phase transactions. In effect we are ignoring the prepare phase and
Remove AELs from subxids correctly on standby Issues relate only to subtransactions that hold AccessExclusiveLocks when replayed on standby. Prior to PG10, aborting subtransactions that held an AccessExclusiveLock failed to release the lock until top level commit or abort. 49bff5300d527 fixed that. However, 49bff5300d527 also introduced a similar bug where subtransaction commit would fail to release an AccessExclusiveLock, leaving the lock to be removed sometimes early and sometimes late. This commit fixes that bug also. Backpatch to PG10 needed. Tested by observation. Note need for multi-node isolationtester to improve test coverage for this and other HS cases. Reported-by: Simon Riggs Author: Simon Riggs
2018-06-16 15:03:29 +02:00
* just going straight to lock release.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
if (parsed->xinfo & XACT_XINFO_HAS_AE_LOCKS)
Remove AELs from subxids correctly on standby Issues relate only to subtransactions that hold AccessExclusiveLocks when replayed on standby. Prior to PG10, aborting subtransactions that held an AccessExclusiveLock failed to release the lock until top level commit or abort. 49bff5300d527 fixed that. However, 49bff5300d527 also introduced a similar bug where subtransaction commit would fail to release an AccessExclusiveLock, leaving the lock to be removed sometimes early and sometimes late. This commit fixes that bug also. Backpatch to PG10 needed. Tested by observation. Note need for multi-node isolationtester to improve test coverage for this and other HS cases. Reported-by: Simon Riggs Author: Simon Riggs
2018-06-16 15:03:29 +02:00
StandbyReleaseLockTree(xid, parsed->nsubxacts, parsed->subxacts);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
Introduce replication progress tracking infrastructure. When implementing a replication solution ontop of logical decoding, two related problems exist: * How to safely keep track of replication progress * How to change replication behavior, based on the origin of a row; e.g. to avoid loops in bi-directional replication setups The solution to these problems, as implemented here, consist out of three parts: 1) 'replication origins', which identify nodes in a replication setup. 2) 'replication progress tracking', which remembers, for each replication origin, how far replay has progressed in a efficient and crash safe manner. 3) The ability to filter out changes performed on the behest of a replication origin during logical decoding; this allows complex replication topologies. E.g. by filtering all replayed changes out. Most of this could also be implemented in "userspace", e.g. by inserting additional rows contain origin information, but that ends up being much less efficient and more complicated. We don't want to require various replication solutions to reimplement logic for this independently. The infrastructure is intended to be generic enough to be reusable. This infrastructure also replaces the 'nodeid' infrastructure of commit timestamps. It is intended to provide all the former capabilities, except that there's only 2^16 different origins; but now they integrate with logical decoding. Additionally more functionality is accessible via SQL. Since the commit timestamp infrastructure has also been introduced in 9.5 (commit 73c986add) changing the API is not a problem. For now the number of origins for which the replication progress can be tracked simultaneously is determined by the max_replication_slots GUC. That GUC is not a perfect match to configure this, but there doesn't seem to be sufficient reason to introduce a separate new one. Bumps both catversion and wal page magic. Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer Discussion: 20150216002155.GI15326@awork2.anarazel.de, 20140923182422.GA15776@alap3.anarazel.de, 20131114172632.GE7522@alap2.anarazel.de
2015-04-29 19:30:53 +02:00
if (parsed->xinfo & XACT_XINFO_HAS_ORIGIN)
{
/* recover apply progress */
replorigin_advance(origin_id, parsed->origin_lsn, lsn,
false /* backward */ , false /* WAL */ );
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* Make sure files supposed to be dropped are dropped */
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (parsed->nrels > 0)
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
{
Update minimum recovery point on truncation. If a file is truncated, we must update minRecoveryPoint. Once a file is truncated, there's no going back; it would not be safe to stop recovery at a point earlier than that anymore. Per report from Kyotaro HORIGUCHI. Backpatch to 8.4. Before that, minRecoveryPoint was not updated during recovery at all.
2012-12-10 14:54:42 +01:00
/*
* First update minimum recovery point to cover this WAL record. Once
* a relation is deleted, there's no going back. The buffer manager
* enforces the WAL-first rule for normal updates to relation files,
* so that the minimum recovery point is always updated before the
* corresponding change in the data file is flushed to disk, but we
* have to do the same here since we're bypassing the buffer manager.
*
* Doing this before deleting the files means that if a deletion fails
* for some reason, you cannot start up the system even after restart,
* until you fix the underlying situation so that the deletion will
* succeed. Alternatively, we could update the minimum recovery point
* after deletion, but that would leave a small window where the
* WAL-first rule would be violated.
*/
XLogFlush(lsn);
Improve the performance of relation deletes during recovery. When multiple relations are deleted at the same transaction, the files of those relations are deleted by one call to smgrdounlinkall(), which leads to scan whole shared_buffers only one time. OTOH, previously, during recovery, smgrdounlink() (not smgrdounlinkall()) was called for each file to delete, which led to scan shared_buffers multiple times. Obviously this could cause to increase the WAL replay time very much especially when shared_buffers was huge. To alleviate this situation, this commit changes the recovery so that it also calls smgrdounlinkall() only one time to delete multiple relation files. This is just fix for oversight of commit 279628a0a7, not new feature. So, per discussion on pgsql-hackers, we concluded to backpatch this to all supported versions. Author: Fujii Masao Reviewed-by: Michael Paquier, Andres Freund, Thomas Munro, Kyotaro Horiguchi, Takayuki Tsunakawa Discussion: https://postgr.es/m/CAHGQGwHVQkdfDqtvGVkty+19cQakAydXn1etGND3X0PHbZ3+6w@mail.gmail.com
2018-07-04 19:21:15 +02:00
/* Make sure files supposed to be dropped are dropped */
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
DropRelationFiles(parsed->xlocators, parsed->nrels, true);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (parsed->nstats > 0)
{
/* see equivalent call for relations above */
XLogFlush(lsn);
pgstat_execute_transactional_drops(parsed->nstats, parsed->stats, true);
}
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* We issue an XLogFlush() for the same reason we emit ForceSyncCommit()
Update minimum recovery point on truncation. If a file is truncated, we must update minRecoveryPoint. Once a file is truncated, there's no going back; it would not be safe to stop recovery at a point earlier than that anymore. Per report from Kyotaro HORIGUCHI. Backpatch to 8.4. Before that, minRecoveryPoint was not updated during recovery at all.
2012-12-10 14:54:42 +01:00
* in normal operation. For example, in CREATE DATABASE, we copy all files
* from the template database, and then commit the transaction. If we
* crash after all the files have been copied but before the commit, you
* have files in the data directory without an entry in pg_database. To
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
* minimize the window for that, we use ForceSyncCommit() to rush the
* commit record to disk as quick as possible. We have the same window
* during recovery, and forcing an XLogFlush() (which updates
* minRecoveryPoint during recovery) helps to reduce that problem window,
* for any user that requested ForceSyncCommit().
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
if (XactCompletionForceSyncCommit(parsed->xinfo))
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
XLogFlush(lsn);
Introduce compact WAL record for the common case of commit (non-DDL). XLOG_XACT_COMMIT_COMPACT leaves out invalidation messages and relfilenodes, saving considerable space for the vast majority of transaction commits. XLOG_XACT_COMMIT keeps same definition as XLOG_PAGE_MAGIC 0xD067 and earlier. Leonardo Francalanci and Simon Riggs
2011-06-28 23:58:17 +02:00
Add new replication mode synchronous_commit = 'remote_apply'. In this mode, the master waits for the transaction to be applied on the remote side, not just written to disk. That means that you can count on a transaction started on the standby to see all commits previously acknowledged by the master. To make this work, the standby sends a reply after replaying each commit record generated with synchronous_commit >= 'remote_apply'. This introduces a small inefficiency: the extra replies will be sent even by standbys that aren't the current synchronous standby. But previously-existing synchronous_commit levels make no attempt at all to optimize which replies are sent based on what the primary cares about, so this is no worse, and at least avoids any extra replies for people not using the feature at all. Thomas Munro, reviewed by Michael Paquier and by me. Some additional tweaks by me.
2016-03-30 03:16:12 +02:00
/*
* If asked by the primary (because someone is waiting for a synchronous
* commit = remote_apply), we will need to ask walreceiver to send a reply
* immediately.
*/
if (XactCompletionApplyFeedback(parsed->xinfo))
XLogRequestWalReceiverReply();
Introduce compact WAL record for the common case of commit (non-DDL). XLOG_XACT_COMMIT_COMPACT leaves out invalidation messages and relfilenodes, saving considerable space for the vast majority of transaction commits. XLOG_XACT_COMMIT keeps same definition as XLOG_PAGE_MAGIC 0xD067 and earlier. Leonardo Francalanci and Simon Riggs
2011-06-28 23:58:17 +02:00
}
Run pgindent on 9.2 source tree in preparation for first 9.3 commit-fest.
2012-06-10 21:20:04 +02:00
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
* Be careful with the order of execution, as with xact_redo_commit().
* The two functions are similar but differ in key places.
*
* Note also that an abort can be for a subtransaction and its children,
* not just for a top level abort. That means we have to consider
* topxid != xid, whereas in commit we would find topxid == xid always
* because subtransaction commit is never WAL logged.
*/
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
static void
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
xact_redo_abort(xl_xact_parsed_abort *parsed, TransactionId xid,
XLogRecPtr lsn, RepOriginId origin_id)
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
{
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionId max_xid;
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
Assert(TransactionIdIsValid(xid));
Rename VariableCacheData.nextFullXid to nextXid. Including Full in variable names duplicates the type information and leads to overly long names. As FullTransactionId cannot accidentally be casted to TransactionId that does not seem necessary. Author: Andres Freund Discussion: https://postgr.es/m/20200724011143.jccsyvsvymuiqfxu@alap3.anarazel.de
2020-08-11 20:25:23 +02:00
/* Make sure nextXid is beyond any XID mentioned in the record. */
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
max_xid = TransactionIdLatest(xid,
parsed->nsubxacts,
parsed->subxacts);
Add basic infrastructure for 64 bit transaction IDs. Instead of inferring epoch progress from xids and checkpoints, introduce a 64 bit FullTransactionId type and use it to track xid generation. This fixes an unlikely bug where the epoch is reported incorrectly if the range of active xids wraps around more than once between checkpoints. The only user-visible effect of this commit is to correct the epoch used by txid_current() and txid_status(), also visible with pg_controldata, in those rare circumstances. It also creates some basic infrastructure so that later patches can use 64 bit transaction IDs in more places. The new type is a struct that we pass by value, as a form of strong typedef. This prevents the sort of accidental confusion between TransactionId and FullTransactionId that would be possible if we were to use a plain old uint64. Author: Thomas Munro Reported-by: Amit Kapila Reviewed-by: Andres Freund, Tom Lane, Heikki Linnakangas Discussion: https://postgr.es/m/CAA4eK1%2BMv%2Bmb0HFfWM9Srtc6MVe160WFurXV68iAFMcagRZ0dQ%40mail.gmail.com
2019-03-27 22:34:43 +01:00
AdvanceNextFullTransactionIdPastXid(max_xid);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
if (standbyState == STANDBY_DISABLED)
{
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
/* Mark the transaction aborted in pg_xact, no need for async stuff */
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionIdAbortTree(xid, parsed->nsubxacts, parsed->subxacts);
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
}
else
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
{
/*
* If a transaction completion record arrives that has as-yet
* unobserved subtransactions then this will not have been fully
* handled by the call to RecordKnownAssignedTransactionIds() in the
* main recovery loop in xlog.c. So we need to do bookkeeping again to
* cover that case. This is confusing and it is easy to think this
* call is irrelevant, which has happened three times in development
* already. Leave it in.
*/
RecordKnownAssignedTransactionIds(max_xid);
Rename "pg_clog" directory to "pg_xact". Names containing the letters "log" sometimes confuse users into believing that only non-critical data is present. It is hoped this renaming will discourage ill-considered removals of transaction status data. Michael Paquier Discussion: http://postgr.es/m/CA+Tgmoa9xFQyjRZupbdEFuwUerFTvC6HjZq1ud6GYragGDFFgA@mail.gmail.com
2017-03-17 14:46:58 +01:00
/* Mark the transaction aborted in pg_xact, no need for async stuff */
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
TransactionIdAbortTree(xid, parsed->nsubxacts, parsed->subxacts);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
* We must update the ProcArray after we have marked clog.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
ExpireTreeKnownAssignedTransactionIds(xid, parsed->nsubxacts, parsed->subxacts, max_xid);
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
/*
Remove leftover reference to old "flat file" mechanism in a comment. The flat file mechanism was removed in PostgreSQL 9.0.
2019-05-08 08:32:34 +02:00
* There are no invalidation messages to send or undo.
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
*/
/*
* Release locks, if any. There are no invalidations to send.
*/
Improve performance of replay of AccessExclusiveLocks A hot standby replica keeps a list of Access Exclusive locks for a top level transaction. These locks are released when the top level transaction ends. Searching of this list is O(N^2), and each transaction had to pay the price of searching this list for locks, even if it didn't take any AE locks itself. This patch optimizes this case by having the master server track which transactions took AE locks, and passes that along to the standby server in the commit/abort record. This allows the standby to only try to release locks for transactions which actually took any, avoiding the majority of the performance issue. Refactor MyXactAccessedTempRel into MyXactFlags to allow minimal additional cruft with this. Analysis and initial patch by David Rowley Author: David Rowley and Simon Riggs
2017-03-22 14:09:36 +01:00
if (parsed->xinfo & XACT_XINFO_HAS_AE_LOCKS)
StandbyReleaseLockTree(xid, parsed->nsubxacts, parsed->subxacts);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
if (parsed->xinfo & XACT_XINFO_HAS_ORIGIN)
{
/* recover apply progress */
replorigin_advance(origin_id, parsed->origin_lsn, lsn,
false /* backward */ , false /* WAL */ );
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
/* Make sure files supposed to be dropped are dropped */
Update minimum recovery point on truncation during WAL replay of abort record. If a file is truncated, we must update minRecoveryPoint. Once a file is truncated, there's no going back; it would not be safe to stop recovery at a point earlier than that anymore. Commit 7bffc9b7bf changed xact_redo_commit() so that it updates minRecoveryPoint on truncation, but forgot to change xact_redo_abort(). Back-patch to all supported versions. Reported-by: mengjuan.cmj@alibaba-inc.com Author: Fujii Masao Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/b029fce3-4fac-4265-968e-16f36ff4d075.mengjuan.cmj@alibaba-inc.com
2021-07-28 18:30:02 +02:00
if (parsed->nrels > 0)
{
/*
* See comments about update of minimum recovery point on truncation,
* in xact_redo_commit().
*/
XLogFlush(lsn);
Change internal RelFileNode references to RelFileNumber or RelFileLocator. We have been using the term RelFileNode to refer to either (1) the integer that is used to name the sequence of files for a certain relation within the directory set aside for that tablespace/database combination; or (2) that value plus the OIDs of the tablespace and database; or occasionally (3) the whole series of files created for a relation based on those values. Using the same name for more than one thing is confusing. Replace RelFileNode with RelFileNumber when we're talking about just the single number, i.e. (1) from above, and with RelFileLocator when we're talking about all the things that are needed to locate a relation's files on disk, i.e. (2) from above. In the places where we refer to (3) as a relfilenode, instead refer to "relation storage". Since there is a ton of SQL code in the world that knows about pg_class.relfilenode, don't change the name of that column, or of other SQL-facing things that derive their name from it. On the other hand, do adjust closely-related internal terminology. For example, the structure member names dbNode and spcNode appear to be derived from the fact that the structure itself was called RelFileNode, so change those to dbOid and spcOid. Likewise, various variables with names like rnode and relnode get renamed appropriately, according to how they're being used in context. Hopefully, this is clearer than before. It is also preparation for future patches that intend to widen the relfilenumber fields from its current width of 32 bits. Variables that store a relfilenumber are now declared as type RelFileNumber rather than type Oid; right now, these are the same, but that can now more easily be changed. Dilip Kumar, per an idea from me. Reviewed also by Andres Freund. I fixed some whitespace issues, changed a couple of words in a comment, and made one other minor correction. Discussion: http://postgr.es/m/CA+TgmoamOtXbVAQf9hWFzonUo6bhhjS6toZQd7HZ-pmojtAmag@mail.gmail.com Discussion: http://postgr.es/m/CA+Tgmobp7+7kmi4gkq7Y+4AM9fTvL+O1oQ4-5gFTT+6Ng-dQ=g@mail.gmail.com Discussion: http://postgr.es/m/CAFiTN-vTe79M8uDH1yprOU64MNFE+R3ODRuA+JWf27JbhY4hJw@mail.gmail.com
2022-07-06 17:39:09 +02:00
DropRelationFiles(parsed->xlocators, parsed->nrels, true);
Update minimum recovery point on truncation during WAL replay of abort record. If a file is truncated, we must update minRecoveryPoint. Once a file is truncated, there's no going back; it would not be safe to stop recovery at a point earlier than that anymore. Commit 7bffc9b7bf changed xact_redo_commit() so that it updates minRecoveryPoint on truncation, but forgot to change xact_redo_abort(). Back-patch to all supported versions. Reported-by: mengjuan.cmj@alibaba-inc.com Author: Fujii Masao Reviewed-by: Heikki Linnakangas Discussion: https://postgr.es/m/b029fce3-4fac-4265-968e-16f36ff4d075.mengjuan.cmj@alibaba-inc.com
2021-07-28 18:30:02 +02:00
}
pgstat: scaffolding for transactional stats creation / drop. One problematic part of the current statistics collector design is that there is no reliable way of getting rid of statistics entries. Because of that pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the current database with the catalog contents and tries to drop now-superfluous entries. That's quite expensive. What's worse, it doesn't work on physical replicas, despite physical replicas collection statistics entries. This commit introduces infrastructure to create / drop statistics entries transactionally, together with the underlying catalog objects (functions, relations, subscriptions). pgstat_xact.c maintains a list of stats entries created / dropped transactionally in the current transaction. To ensure the removal of statistics entries is durable dropped statistics entries are included in commit / abort (and prepare) records, which also ensures that stats entries are dropped on standbys. Statistics entries created separately from creating the underlying catalog object (e.g. when stats were previously lost due to an immediate restart) are *not* WAL logged. However that can only happen outside of the transaction creating the catalog object, so it does not lead to "leaked" statistics entries. For this to work, functions creating / dropping functions / relations / subscriptions need to call into pgstat. For subscriptions this was already done when dropping subscriptions, via pgstat_report_subscription_drop() (now renamed to pgstat_drop_subscription()). This commit does not actually drop stats yet, it just provides the infrastructure. It is however a largely independent piece of infrastructure, so committing it separately makes sense. Bumps XLOG_PAGE_MAGIC. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de
2022-04-07 03:22:22 +02:00
if (parsed->nstats > 0)
{
/* see equivalent call for relations above */
XLogFlush(lsn);
pgstat_execute_transactional_drops(parsed->nstats, parsed->stats, true);
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
void
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
xact_redo(XLogReaderState *record)
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
{
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
uint8 info = XLogRecGetInfo(record) & XLOG_XACT_OPMASK;
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
Add a new option to RestoreBkpBlocks() to indicate if a cleanup lock should be used instead of the normal exclusive lock, and make WAL redo functions responsible for calling RestoreBkpBlocks(). They know better what kind of a lock they need. At the moment, this just moves things around with no functional change, but makes the hot standby patch that's under review cleaner.
2009-01-20 19:59:37 +01:00
/* Backup blocks are not used in xact records */
Revamp the WAL record format. Each WAL record now carries information about the modified relation and block(s) in a standardized format. That makes it easier to write tools that need that information, like pg_rewind, prefetching the blocks to speed up recovery, etc. There's a whole new API for building WAL records, replacing the XLogRecData chains used previously. The new API consists of XLogRegister* functions, which are called for each buffer and chunk of data that is added to the record. The new API also gives more control over when a full-page image is written, by passing flags to the XLogRegisterBuffer function. This also simplifies the XLogReadBufferForRedo() calls. The function can dig the relation and block number from the WAL record, so they no longer need to be passed as arguments. For the convenience of redo routines, XLogReader now disects each WAL record after reading it, copying the main data part and the per-block data into MAXALIGNed buffers. The data chunks are not aligned within the WAL record, but the redo routines can assume that the pointers returned by XLogRecGet* functions are. Redo routines are now passed the XLogReaderState, which contains the record in the already-disected format, instead of the plain XLogRecord. The new record format also makes the fixed size XLogRecord header smaller, by removing the xl_len field. The length of the "main data" portion is now stored at the end of the WAL record, and there's a separate header after XLogRecord for it. The alignment padding at the end of XLogRecord is also removed. This compansates for the fact that the new format would otherwise be more bulky than the old format. Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera, Fujii Masao.
2014-11-20 16:56:26 +01:00
Assert(!XLogRecHasAnyBlockRefs(record));
Add a new option to RestoreBkpBlocks() to indicate if a cleanup lock should be used instead of the normal exclusive lock, and make WAL redo functions responsible for calling RestoreBkpBlocks(). They know better what kind of a lock they need. At the moment, this just moves things around with no functional change, but makes the hot standby patch that's under review cleaner.
2009-01-20 19:59:37 +01:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
if (info == XLOG_XACT_COMMIT)
redo/undo support functions and cleanups.
2000-10-20 13:01:21 +02:00
{
Commit the reasonably uncontroversial parts of J.R. Nield's PITR patch, to wit: Add a header record to each WAL segment file so that it can be reliably identified. Avoid splitting WAL records across segment files (this is not strictly necessary, but makes it simpler to incorporate the header records). Make WAL entries for file creation, deletion, and truncation (as foreseen but never implemented by Vadim). Also, add support for making XLOG_SEG_SIZE configurable at compile time, similarly to BLCKSZ. Fix a couple bugs I introduced in WAL replay during recent smgr API changes. initdb is forced due to changes in pg_control contents.
2004-02-11 23:55:26 +01:00
xl_xact_commit *xlrec = (xl_xact_commit *) XLogRecGetData(record);
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xl_xact_parsed_commit parsed;
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
ParseCommitRecord(XLogRecGetInfo(record), xlrec, &parsed);
xact_redo_commit(&parsed, XLogRecGetXid(record),
record->EndRecPtr, XLogRecGetOrigin(record));
}
else if (info == XLOG_XACT_COMMIT_PREPARED)
{
xl_xact_commit *xlrec = (xl_xact_commit *) XLogRecGetData(record);
xl_xact_parsed_commit parsed;
Commit the reasonably uncontroversial parts of J.R. Nield's PITR patch, to wit: Add a header record to each WAL segment file so that it can be reliably identified. Avoid splitting WAL records across segment files (this is not strictly necessary, but makes it simpler to incorporate the header records). Make WAL entries for file creation, deletion, and truncation (as foreseen but never implemented by Vadim). Also, add support for making XLOG_SEG_SIZE configurable at compile time, similarly to BLCKSZ. Fix a couple bugs I introduced in WAL replay during recent smgr API changes. initdb is forced due to changes in pg_control contents.
2004-02-11 23:55:26 +01:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
ParseCommitRecord(XLogRecGetInfo(record), xlrec, &parsed);
xact_redo_commit(&parsed, parsed.twophase_xid,
record->EndRecPtr, XLogRecGetOrigin(record));
Speedup 2PC recovery by skipping two phase state files in normal path 2PC state info held in shmem at PREPARE, then cleaned at COMMIT PREPARED/ABORT PREPARED, avoiding writing/fsyncing any state information to disk in the normal path, greatly enhancing replay speed. Prepared transactions that live past one checkpoint redo horizon will be written to disk as now. Similar conceptually to 978b2f65aa1262eb4ecbf8b3785cb1b9cf4db78e and building upon the infrastructure created by that commit. Authors, in equal measure: Stas Kelvich, Nikhil Sontakke and Michael Paquier Discussion: https://postgr.es/m/CAMGcDxf8Bn9ZPBBJZba9wiyQq-Qk5uqq=VjoMnRnW5s+fKST3w@mail.gmail.com
2017-04-04 21:56:56 +02:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
/* Delete TwoPhaseState gxact entry and/or 2PC file. */
LWLockAcquire(TwoPhaseStateLock, LW_EXCLUSIVE);
PrepareRedoRemove(parsed.twophase_xid, false);
LWLockRelease(TwoPhaseStateLock);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
else if (info == XLOG_XACT_ABORT)
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
{
xl_xact_abort *xlrec = (xl_xact_abort *) XLogRecGetData(record);
Merge the various forms of transaction commit & abort records. Since 465883b0a two versions of commit records have existed. A compact version that was used when no cache invalidations, smgr unlinks and similar were needed, and a full version that could deal with all that. Additionally the full version was embedded into twophase commit records. That resulted in a measurable reduction in the size of the logged WAL in some workloads. But more recently additions like logical decoding, which e.g. needs information about the database something was executed on, made it applicable in fewer situations. The static split generally made it hard to expand the commit record, because concerns over the size made it hard to add anything to the compact version. Additionally it's not particularly pretty to have twophase.c insert RM_XACT records. Rejigger things so that the commit and abort records only have one form each, including the twophase equivalents. The presence of the various optional (in the sense of not being in every record) pieces is indicated by a bits in the 'xinfo' flag. That flag previously was not included in compact commit records. To prevent an increase in size due to its presence, it's only included if necessary; signalled by a bit in the xl_info bits available for xact.c, similar to heapam.c's XLOG_HEAP_OPMASK/XLOG_HEAP_INIT_PAGE. Twophase commit/aborts are now the same as their normal counterparts. The original transaction's xid is included in an optional data field. This means that commit records generally are smaller, except in the case of a transaction with subtransactions, but no other special cases; the increase there is four bytes, which seems acceptable given that the more common case of not having subtransactions shrank. The savings are especially measurable for twophase commits, which previously always used the full version; but will in practice only infrequently have required that. The motivation for this work are not the space savings and and deduplication though; it's that it makes it easier to extend commit records with additional information. That's just a few lines of code now; without impacting the common case where that information is not needed. Discussion: 20150220152150.GD4149@awork2.anarazel.de, 235610.92468.qm%40web29004.mail.ird.yahoo.com Reviewed-By: Heikki Linnakangas, Simon Riggs
2015-03-15 17:37:07 +01:00
xl_xact_parsed_abort parsed;
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
ParseAbortRecord(XLogRecGetInfo(record), xlrec, &parsed);
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
xact_redo_abort(&parsed, XLogRecGetXid(record),
record->EndRecPtr, XLogRecGetOrigin(record));
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
}
else if (info == XLOG_XACT_ABORT_PREPARED)
{
xl_xact_abort *xlrec = (xl_xact_abort *) XLogRecGetData(record);
xl_xact_parsed_abort parsed;
Fix failure to advance nextXID beyond subtransactions whose XIDs appear only within COMMIT or ABORT records.
2004-08-30 21:00:03 +02:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
ParseAbortRecord(XLogRecGetInfo(record), xlrec, &parsed);
Track replication origin progress for rollbacks. Commit 1eb6d6527a allowed to track replica origin replay progress for 2PC but it was not complete. It misses to properly track the progress for rollback prepared especially it missed updating the code for recovery. Additionally, we need to allow tracking it on subscriber nodes where wal_level might not be logical. It is required to track decoding of 2PC which is committed in PG14 (a271a1b50e) and also nobody complained about this till now so not backpatching it. Author: Amit Kapila Reviewed-by: Michael Paquier and Ajin Cherian Discussion: https://postgr.es/m/CAA4eK1L-kHmMnSdrRW6UhRbCjR7cgh04c+6psY15qzT6ktcd+g@mail.gmail.com
2021-03-08 03:24:03 +01:00
xact_redo_abort(&parsed, parsed.twophase_xid,
record->EndRecPtr, XLogRecGetOrigin(record));
Speedup 2PC recovery by skipping two phase state files in normal path 2PC state info held in shmem at PREPARE, then cleaned at COMMIT PREPARED/ABORT PREPARED, avoiding writing/fsyncing any state information to disk in the normal path, greatly enhancing replay speed. Prepared transactions that live past one checkpoint redo horizon will be written to disk as now. Similar conceptually to 978b2f65aa1262eb4ecbf8b3785cb1b9cf4db78e and building upon the infrastructure created by that commit. Authors, in equal measure: Stas Kelvich, Nikhil Sontakke and Michael Paquier Discussion: https://postgr.es/m/CAMGcDxf8Bn9ZPBBJZba9wiyQq-Qk5uqq=VjoMnRnW5s+fKST3w@mail.gmail.com
2017-04-04 21:56:56 +02:00
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
/* Delete TwoPhaseState gxact entry and/or 2PC file. */
LWLockAcquire(TwoPhaseStateLock, LW_EXCLUSIVE);
PrepareRedoRemove(parsed.twophase_xid, false);
LWLockRelease(TwoPhaseStateLock);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
else if (info == XLOG_XACT_PREPARE)
{
Speedup 2PC recovery by skipping two phase state files in normal path 2PC state info held in shmem at PREPARE, then cleaned at COMMIT PREPARED/ABORT PREPARED, avoiding writing/fsyncing any state information to disk in the normal path, greatly enhancing replay speed. Prepared transactions that live past one checkpoint redo horizon will be written to disk as now. Similar conceptually to 978b2f65aa1262eb4ecbf8b3785cb1b9cf4db78e and building upon the infrastructure created by that commit. Authors, in equal measure: Stas Kelvich, Nikhil Sontakke and Michael Paquier Discussion: https://postgr.es/m/CAMGcDxf8Bn9ZPBBJZba9wiyQq-Qk5uqq=VjoMnRnW5s+fKST3w@mail.gmail.com
2017-04-04 21:56:56 +02:00
/*
* Store xid and start/end pointers of the WAL record in TwoPhaseState
* gxact entry.
*/
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
LWLockAcquire(TwoPhaseStateLock, LW_EXCLUSIVE);
Speedup 2PC recovery by skipping two phase state files in normal path 2PC state info held in shmem at PREPARE, then cleaned at COMMIT PREPARED/ABORT PREPARED, avoiding writing/fsyncing any state information to disk in the normal path, greatly enhancing replay speed. Prepared transactions that live past one checkpoint redo horizon will be written to disk as now. Similar conceptually to 978b2f65aa1262eb4ecbf8b3785cb1b9cf4db78e and building upon the infrastructure created by that commit. Authors, in equal measure: Stas Kelvich, Nikhil Sontakke and Michael Paquier Discussion: https://postgr.es/m/CAMGcDxf8Bn9ZPBBJZba9wiyQq-Qk5uqq=VjoMnRnW5s+fKST3w@mail.gmail.com
2017-04-04 21:56:56 +02:00
PrepareRedoAdd(XLogRecGetData(record),
record->ReadRecPtr,
Store 2PC GID in commit/abort WAL recs for logical decoding Store GID of 2PC in commit/abort WAL records when wal_level = logical. This allows logical decoding to send the SAME gid to subscribers across restarts of logical replication. Track relica origin replay progress for 2PC. (Edited from patch 0003 in the logical decoding 2PC series.) Authors: Nikhil Sontakke, Stas Kelvich Reviewed-by: Simon Riggs, Andres Freund
2018-03-28 18:42:50 +02:00
record->EndRecPtr,
XLogRecGetOrigin(record));
Avoid bogus TwoPhaseState locking sequences The optimized code in 728bd991c3c4 contains a few invalid locking sequences. To wit, the original code would try to acquire an lwlock that it already holds. Avoid this by moving lock acquisitions to higher-level code, and install appropriate assertions in low-level that the correct mode is held. Authors: Michael Paquier, Álvaro Herrera Reported-By: chuanting wang Bug: #14680 Discussion: https://postgr.es/m/20170531033228.1487.10124@wrigleys.postgresql.org
2017-06-14 17:29:05 +02:00
LWLockRelease(TwoPhaseStateLock);
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
}
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
else if (info == XLOG_XACT_ASSIGNMENT)
{
xl_xact_assignment *xlrec = (xl_xact_assignment *) XLogRecGetData(record);
Cleanup initialization of Hot Standby. Clarify working with reanalysis of requirements and documentation on LogStandbySnapshot(). Fixes two minor bugs reported by Tom Lane that would lead to an incorrect snapshot after transaction wraparound. Also fix two other problems discovered that would give incorrect snapshots in certain cases. ProcArrayApplyRecoveryInfo() substantially rewritten. Some minor refactoring of xact_redo_apply() and ExpireTreeKnownAssignedTransactionIds().
2010-05-13 13:15:38 +02:00
if (standbyState >= STANDBY_INITIALIZED)
Allow read only connections during recovery, known as Hot Standby. Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record. New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far. This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required. Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit. Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
2009-12-19 02:32:45 +01:00
ProcArrayApplyXidAssignment(xlrec->xtop,
xlrec->nsubxacts, xlrec->xsub);
}
WAL Log invalidations at command end with wal_level=logical. When wal_level=logical, write invalidations at command end into WAL so that decoding can use this information. This patch is required to allow the streaming of in-progress transactions in logical decoding.  The actual work to allow streaming will be committed as a separate patch. We still add the invalidations to the cache and write them to WAL at commit time in RecordTransactionCommit(). This uses the existing XLOG_INVALIDATIONS xlog record type, from the RM_STANDBY_ID resource manager (see LogStandbyInvalidations for details). So existing code relying on those invalidations (e.g. redo) does not need to be changed. The invalidations written at command end uses a new xlog record type XLOG_XACT_INVALIDATIONS, from RM_XACT_ID resource manager. See LogLogicalInvalidations for details. These new xlog records are ignored by existing redo procedures, which still rely on the invalidations written to commit records. The invalidations are decoded and accumulated in top-transaction, and then executed during replay.  This obviates the need to decode the invalidations as part of a commit record. Bump XLOG_PAGE_MAGIC, since this introduces XLOG_XACT_INVALIDATIONS. Author: Dilip Kumar, Tomas Vondra, Amit Kapila Reviewed-by: Amit Kapila Tested-by: Neha Sharma and Mahendra Singh Thalor Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2020-07-23 04:49:07 +02:00
else if (info == XLOG_XACT_INVALIDATIONS)
{
/*
* XXX we do ignore this for now, what matters are invalidations
* written into the commit record.
*/
}
Two-phase commit. Original patch by Heikki Linnakangas, with additional hacking by Alvaro Herrera and Tom Lane.
2005-06-18 00:32:51 +02:00
else
elog(PANIC, "xact_redo: unknown op code %u", info);
}