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/replication/logical/reorderbuffer.c

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

5281 lines
152 KiB
C
Raw Normal View History

Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*-------------------------------------------------------------------------
*
* reorderbuffer.c
* PostgreSQL logical replay/reorder buffer management
*
*
Update copyright for 2023 Backpatch-through: 11
2023-01-02 21:00:37 +01:00
* Copyright (c) 2012-2023, PostgreSQL Global Development Group
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*
*
* IDENTIFICATION
Fix various typos, grammar and code style in comments and docs This fixes a set of issues that have accumulated over the past months (or years) in various code areas. Most fixes are related to some recent additions, as of the development of v15. Author: Justin Pryzby Discussion: https://postgr.es/m/20220124030001.GQ23027@telsasoft.com
2022-01-25 01:40:04 +01:00
* src/backend/replication/logical/reorderbuffer.c
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*
* NOTES
* This module gets handed individual pieces of transactions in the order
* they are written to the WAL and is responsible to reassemble them into
* toplevel transaction sized pieces. When a transaction is completely
Spelling adjustments
2020-06-07 15:06:51 +02:00
* reassembled - signaled by reading the transaction commit record - it
Fix Latin spelling "c.f." should be "cf.".
2018-01-11 14:31:11 +01:00
* will then call the output plugin (cf. ReorderBufferCommit()) with the
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* individual changes. The output plugins rely on snapshots built by
* snapbuild.c which hands them to us.
*
* Transactions and subtransactions/savepoints in postgres are not
* immediately linked to each other from outside the performing
* backend. Only at commit/abort (or special xact_assignment records) they
* are linked together. Which means that we will have to splice together a
* toplevel transaction from its subtransactions. To do that efficiently we
* build a binary heap indexed by the smallest current lsn of the individual
* subtransactions' changestreams. As the individual streams are inherently
* ordered by LSN - since that is where we build them from - the transaction
* can easily be reassembled by always using the subtransaction with the
* smallest current LSN from the heap.
*
* In order to cope with large transactions - which can be several times as
* big as the available memory - this module supports spooling the contents
* of a large transactions to disk. When the transaction is replayed the
* contents of individual (sub-)transactions will be read from disk in
* chunks.
*
* This module also has to deal with reassembling toast records from the
* individual chunks stored in WAL. When a new (or initial) version of a
* tuple is stored in WAL it will always be preceded by the toast chunks
* emitted for the columns stored out of line. Within a single toplevel
* transaction there will be no other data carrying records between a row's
* toast chunks and the row data itself. See ReorderBufferToast* for
* details.
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
*
* ReorderBuffer uses two special memory context types - SlabContext for
* allocations of fixed-length structures (changes and transactions), and
* GenerationContext for the variable-length transaction data (allocated
Fix comments in reorderbuffer.c. Author: Dave Cramer Reviewed-by: David G. Johnston Discussion: https://postgr.es/m/CADK3HHL8do4Fp1bsymgNasx375njV3AR7zY3UgYwzbL_Dx-n2Q@mail.gmail.com
2020-07-18 06:17:38 +02:00
* and freed in groups with similar lifespans).
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
*
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
* To limit the amount of memory used by decoded changes, we track memory
* used at the reorder buffer level (i.e. total amount of memory), and for
* each transaction. When the total amount of used memory exceeds the
* limit, the transaction consuming the most memory is then serialized to
* disk.
*
* Only decoded changes are evicted from memory (spilled to disk), not the
* transaction records. The number of toplevel transactions is limited,
* but a transaction with many subtransactions may still consume significant
Adjust minor comment in reorderbuffer.c. Author: Dave Cramer Reviewed-by: David G. Johnston Discussion: https://postgr.es/m/CADK3HHL8do4Fp1bsymgNasx375njV3AR7zY3UgYwzbL_Dx-n2Q@mail.gmail.com
2020-07-18 06:27:23 +02:00
* amounts of memory. However, the transaction records are fairly small and
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
* are not included in the memory limit.
*
* The current eviction algorithm is very simple - the transaction is
* picked merely by size, while it might be useful to also consider age
* (LSN) of the changes for example. With the new Generational memory
* allocator, evicting the oldest changes would make it more likely the
* memory gets actually freed.
*
* We still rely on max_changes_in_memory when loading serialized changes
* back into memory. At that point we can't use the memory limit directly
Fix comments in reorderbuffer.c. Author: Dave Cramer Reviewed-by: David G. Johnston Discussion: https://postgr.es/m/CADK3HHL8do4Fp1bsymgNasx375njV3AR7zY3UgYwzbL_Dx-n2Q@mail.gmail.com
2020-07-18 06:17:38 +02:00
* as we load the subxacts independently. One option to deal with this
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
* would be to count the subxacts, and allow each to allocate 1/N of the
* memory limit. That however does not seem very appealing, because with
Fix comments in reorderbuffer.c. Author: Dave Cramer Reviewed-by: David G. Johnston Discussion: https://postgr.es/m/CADK3HHL8do4Fp1bsymgNasx375njV3AR7zY3UgYwzbL_Dx-n2Q@mail.gmail.com
2020-07-18 06:17:38 +02:00
* many subtransactions it may easily cause thrashing (short cycles of
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
* deserializing and applying very few changes). We probably should give
* a bit more memory to the oldest subtransactions, because it's likely
Fix comments in reorderbuffer.c. Author: Dave Cramer Reviewed-by: David G. Johnston Discussion: https://postgr.es/m/CADK3HHL8do4Fp1bsymgNasx375njV3AR7zY3UgYwzbL_Dx-n2Q@mail.gmail.com
2020-07-18 06:17:38 +02:00
* they are the source for the next sequence of changes.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* -------------------------------------------------------------------------
*/
#include "postgres.h"
#include <unistd.h>
#include <sys/stat.h>
Split tuptoaster.c into three separate files. detoast.c/h contain functions required to detoast a datum, partially or completely, plus a few other utility functions for examining the size of toasted datums. toast_internals.c/h contain functions that are used internally to the TOAST subsystem but which (mostly) do not need to be accessed from outside. heaptoast.c/h contains code that is intrinsically specific to the heap AM, either because it operates on HeapTuples or is based on the layout of a heap page. detoast.c and toast_internals.c are placed in src/backend/access/common rather than src/backend/access/heap. At present, both files still have dependencies on the heap, but that will be improved in a future commit. Patch by me, reviewed and tested by Prabhat Sabu, Thomas Munro, Andres Freund, and Álvaro Herrera. Discussion: http://postgr.es/m/CA+TgmoZv-=2iWM4jcw5ZhJeL18HF96+W1yJeYrnGMYdkFFnEpQ@mail.gmail.com
2019-07-08 17:58:05 +02:00
#include "access/detoast.h"
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
#include "access/heapam.h"
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
#include "access/rewriteheap.h"
#include "access/transam.h"
#include "access/xact.h"
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
#include "access/xlog_internal.h"
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
#include "catalog/catalog.h"
#include "lib/binaryheap.h"
Rationalize common/relpath.[hc]. Commit a73018392636ce832b09b5c31f6ad1f18a4643ea created rather a mess by putting dependencies on backend-only include files into include/common. We really shouldn't do that. To clean it up: * Move TABLESPACE_VERSION_DIRECTORY back to its longtime home in catalog/catalog.h. We won't consider this symbol part of the FE/BE API. * Push enum ForkNumber from relfilenode.h into relpath.h. We'll consider relpath.h as the source of truth for fork numbers, since relpath.c was already partially serving that function, and anyway relfilenode.h was kind of a random place for that enum. * So, relfilenode.h now includes relpath.h rather than vice-versa. This direction of dependency is fine. (That allows most, but not quite all, of the existing explicit #includes of relpath.h to go away again.) * Push forkname_to_number from catalog.c to relpath.c, just to centralize fork number stuff a bit better. * Push GetDatabasePath from catalog.c to relpath.c; it was rather odd that the previous commit didn't keep this together with relpath(). * To avoid needing relfilenode.h in common/, redefine the underlying function (now called GetRelationPath) as taking separate OID arguments, and make the APIs using RelFileNode or RelFileNodeBackend into macro wrappers. (The macros have a potential multiple-eval risk, but none of the existing call sites have an issue with that; one of them had such a risk already anyway.) * Fix failure to follow the directions when "init" fork type was added; specifically, the errhint in forkname_to_number wasn't updated, and neither was the SGML documentation for pg_relation_size(). * Fix tablespace-path-too-long check in CreateTableSpace() to account for fork-name component of maximum-length pathnames. This requires putting FORKNAMECHARS into a header file, but it was rather useless (and actually unreferenced) where it was. The last couple of items are potentially back-patchable bug fixes, if anyone is sufficiently excited about them; but personally I'm not. Per a gripe from Christoph Berg about how include/common wasn't self-contained.
2014-04-30 23:30:50 +02:00
#include "miscadmin.h"
Create and use wait events for read, write, and fsync operations. Previous commits, notably 53be0b1add7064ca5db3cd884302dfc3268d884e and 6f3bd98ebfc008cbd676da777bb0b2376c4c4bfa, made it possible to see from pg_stat_activity when a backend was stuck waiting for another backend, but it's also fairly common for a backend to be stuck waiting for an I/O. Add wait events for those operations, too. Rushabh Lathia, with further hacking by me. Reviewed and tested by Michael Paquier, Amit Kapila, Rajkumar Raghuwanshi, and Rahila Syed. Discussion: http://postgr.es/m/CAGPqQf0LsYHXREPAZqYGVkDqHSyjf=KsD=k0GTVPAuzyThh-VQ@mail.gmail.com
2017-03-18 12:43:01 +01:00
#include "pgstat.h"
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
#include "replication/logical.h"
#include "replication/reorderbuffer.h"
#include "replication/slot.h"
#include "replication/snapbuild.h" /* just for SnapBuildSnapDecRefcount */
#include "storage/bufmgr.h"
#include "storage/fd.h"
#include "storage/sinval.h"
#include "utils/builtins.h"
#include "utils/combocid.h"
#include "utils/memdebug.h"
#include "utils/memutils.h"
Don't include rel.h when relcache.h is sufficient Trivial change to reduce exposure of rel.h.
2015-08-11 18:03:14 +02:00
#include "utils/rel.h"
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
#include "utils/relfilenumbermap.h"
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Rationalize common/relpath.[hc]. Commit a73018392636ce832b09b5c31f6ad1f18a4643ea created rather a mess by putting dependencies on backend-only include files into include/common. We really shouldn't do that. To clean it up: * Move TABLESPACE_VERSION_DIRECTORY back to its longtime home in catalog/catalog.h. We won't consider this symbol part of the FE/BE API. * Push enum ForkNumber from relfilenode.h into relpath.h. We'll consider relpath.h as the source of truth for fork numbers, since relpath.c was already partially serving that function, and anyway relfilenode.h was kind of a random place for that enum. * So, relfilenode.h now includes relpath.h rather than vice-versa. This direction of dependency is fine. (That allows most, but not quite all, of the existing explicit #includes of relpath.h to go away again.) * Push forkname_to_number from catalog.c to relpath.c, just to centralize fork number stuff a bit better. * Push GetDatabasePath from catalog.c to relpath.c; it was rather odd that the previous commit didn't keep this together with relpath(). * To avoid needing relfilenode.h in common/, redefine the underlying function (now called GetRelationPath) as taking separate OID arguments, and make the APIs using RelFileNode or RelFileNodeBackend into macro wrappers. (The macros have a potential multiple-eval risk, but none of the existing call sites have an issue with that; one of them had such a risk already anyway.) * Fix failure to follow the directions when "init" fork type was added; specifically, the errhint in forkname_to_number wasn't updated, and neither was the SGML documentation for pg_relation_size(). * Fix tablespace-path-too-long check in CreateTableSpace() to account for fork-name component of maximum-length pathnames. This requires putting FORKNAMECHARS into a header file, but it was rather useless (and actually unreferenced) where it was. The last couple of items are potentially back-patchable bug fixes, if anyone is sufficiently excited about them; but personally I'm not. Per a gripe from Christoph Berg about how include/common wasn't self-contained.
2014-04-30 23:30:50 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* entry for a hash table we use to map from xid to our transaction state */
typedef struct ReorderBufferTXNByIdEnt
{
TransactionId xid;
ReorderBufferTXN *txn;
} ReorderBufferTXNByIdEnt;
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
/* data structures for (relfilelocator, ctid) => (cmin, cmax) mapping */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
typedef struct ReorderBufferTupleCidKey
{
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 rlocator;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ItemPointerData tid;
} ReorderBufferTupleCidKey;
typedef struct ReorderBufferTupleCidEnt
{
ReorderBufferTupleCidKey key;
CommandId cmin;
CommandId cmax;
CommandId combocid; /* just for debugging */
} ReorderBufferTupleCidEnt;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
/* Virtual file descriptor with file offset tracking */
typedef struct TXNEntryFile
{
File vfd; /* -1 when the file is closed */
off_t curOffset; /* offset for next write or read. Reset to 0
* when vfd is opened. */
} TXNEntryFile;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* k-way in-order change iteration support structures */
typedef struct ReorderBufferIterTXNEntry
{
XLogRecPtr lsn;
ReorderBufferChange *change;
ReorderBufferTXN *txn;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
TXNEntryFile file;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
XLogSegNo segno;
} ReorderBufferIterTXNEntry;
typedef struct ReorderBufferIterTXNState
{
binaryheap *heap;
Size nr_txns;
dlist_head old_change;
ReorderBufferIterTXNEntry entries[FLEXIBLE_ARRAY_MEMBER];
} ReorderBufferIterTXNState;
/* toast datastructures */
typedef struct ReorderBufferToastEnt
{
Oid chunk_id; /* toast_table.chunk_id */
int32 last_chunk_seq; /* toast_table.chunk_seq of the last chunk we
* have seen */
Size num_chunks; /* number of chunks we've already seen */
Size size; /* combined size of chunks seen */
dlist_head chunks; /* linked list of chunks */
struct varlena *reconstructed; /* reconstructed varlena now pointed to in
* main tup */
} ReorderBufferToastEnt;
/* Disk serialization support datastructures */
typedef struct ReorderBufferDiskChange
{
Size size;
ReorderBufferChange change;
/* data follows */
} ReorderBufferDiskChange;
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
#define IsSpecInsert(action) \
( \
((action) == REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT) \
)
Allow streaming the changes after speculative aborts. Until now, we didn't allow to stream the changes in logical replication till we receive speculative confirm or the next DML change record after speculative inserts. The reason was that we never use to process speculative aborts but after commit 4daa140a2f it is possible to process them so we can allow streaming once we receive speculative abort after speculative insertion. We decided to backpatch to 14 where the feature for streaming in progress transactions have been introduced as this is a minor change and makes that functionality better. Author: Amit Kapila Reviewed-By: Dilip Kumar Backpatch-through: 14 Discussion: https://postgr.es/m/CAA4eK1KdqmTCtrBR6oFfGELrLLbDLDedL6zACcsUOQuTJBj1vw@mail.gmail.com
2021-06-30 06:07:59 +02:00
#define IsSpecConfirmOrAbort(action) \
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
( \
Allow streaming the changes after speculative aborts. Until now, we didn't allow to stream the changes in logical replication till we receive speculative confirm or the next DML change record after speculative inserts. The reason was that we never use to process speculative aborts but after commit 4daa140a2f it is possible to process them so we can allow streaming once we receive speculative abort after speculative insertion. We decided to backpatch to 14 where the feature for streaming in progress transactions have been introduced as this is a minor change and makes that functionality better. Author: Amit Kapila Reviewed-By: Dilip Kumar Backpatch-through: 14 Discussion: https://postgr.es/m/CAA4eK1KdqmTCtrBR6oFfGELrLLbDLDedL6zACcsUOQuTJBj1vw@mail.gmail.com
2021-06-30 06:07:59 +02:00
(((action) == REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM) || \
((action) == REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT)) \
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
)
#define IsInsertOrUpdate(action) \
( \
(((action) == REORDER_BUFFER_CHANGE_INSERT) || \
((action) == REORDER_BUFFER_CHANGE_UPDATE) || \
((action) == REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT)) \
)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Maximum number of changes kept in memory, per transaction. After that,
* changes are spooled to disk.
*
* The current value should be sufficient to decode the entire transaction
* without hitting disk in OLTP workloads, while starting to spool to disk in
* other workloads reasonably fast.
*
Fix more typos in comments. Patch by CharSyam, plus a few more I spotted with grep.
2015-05-20 18:44:46 +02:00
* At some point in the future it probably makes sense to have a more elaborate
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* resource management here, but it's not entirely clear what that would look
* like.
*/
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
int logical_decoding_work_mem;
static const Size max_changes_in_memory = 4096; /* XXX for restore only */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
/* GUC variable */
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
int logical_replication_mode = LOGICAL_REP_MODE_BUFFERED;
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* ---------------------------------------
* primary reorderbuffer support routines
* ---------------------------------------
*/
static ReorderBufferTXN *ReorderBufferGetTXN(ReorderBuffer *rb);
static void ReorderBufferReturnTXN(ReorderBuffer *rb, ReorderBufferTXN *txn);
static ReorderBufferTXN *ReorderBufferTXNByXid(ReorderBuffer *rb,
TransactionId xid, bool create, bool *is_new,
XLogRecPtr lsn, bool create_as_top);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
static void ReorderBufferTransferSnapToParent(ReorderBufferTXN *txn,
ReorderBufferTXN *subtxn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void AssertTXNLsnOrder(ReorderBuffer *rb);
/* ---------------------------------------
* support functions for lsn-order iterating over the ->changes of a
* transaction and its subtransactions
*
* used for iteration over the k-way heap merge of a transaction and its
* subtransactions
* ---------------------------------------
*/
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
static void ReorderBufferIterTXNInit(ReorderBuffer *rb, ReorderBufferTXN *txn,
ReorderBufferIterTXNState *volatile *iter_state);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static ReorderBufferChange *ReorderBufferIterTXNNext(ReorderBuffer *rb, ReorderBufferIterTXNState *state);
static void ReorderBufferIterTXNFinish(ReorderBuffer *rb,
ReorderBufferIterTXNState *state);
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
static void ReorderBufferExecuteInvalidations(uint32 nmsgs, SharedInvalidationMessage *msgs);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* ---------------------------------------
* Disk serialization support functions
* ---------------------------------------
*/
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
static void ReorderBufferCheckMemoryLimit(ReorderBuffer *rb);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void ReorderBufferSerializeTXN(ReorderBuffer *rb, ReorderBufferTXN *txn);
static void ReorderBufferSerializeChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
int fd, ReorderBufferChange *change);
static Size ReorderBufferRestoreChanges(ReorderBuffer *rb, ReorderBufferTXN *txn,
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
TXNEntryFile *file, XLogSegNo *segno);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void ReorderBufferRestoreChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
Harmonize reorderbuffer parameter names. Make reorderbuffer.h function declarations consistently use named parameters. Also make sure that the declarations use names that match corresponding names from function definitions in reorderbuffer.c. This makes the definitions easier to follow, especially in the case of functions that happen to have adjoining arguments of the same type. This patch was written with help from clang-tidy. Specifically, its "readability-inconsistent-declaration-parameter-name" check and its "readability-named-parameter" check were used. Author: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/3955318.1663377656@sss.pgh.pa.us
2022-09-18 02:20:17 +02:00
char *data);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void ReorderBufferRestoreCleanup(ReorderBuffer *rb, ReorderBufferTXN *txn);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
static void ReorderBufferTruncateTXN(ReorderBuffer *rb, ReorderBufferTXN *txn,
bool txn_prepared);
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
static void ReorderBufferCleanupSerializedTXNs(const char *slotname);
static void ReorderBufferSerializedPath(char *path, ReplicationSlot *slot,
TransactionId xid, XLogSegNo segno);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void ReorderBufferFreeSnap(ReorderBuffer *rb, Snapshot snap);
static Snapshot ReorderBufferCopySnap(ReorderBuffer *rb, Snapshot orig_snap,
ReorderBufferTXN *txn, CommandId cid);
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
/*
* ---------------------------------------
* Streaming support functions
* ---------------------------------------
*/
static inline bool ReorderBufferCanStream(ReorderBuffer *rb);
static inline bool ReorderBufferCanStartStreaming(ReorderBuffer *rb);
static void ReorderBufferStreamTXN(ReorderBuffer *rb, ReorderBufferTXN *txn);
static void ReorderBufferStreamCommit(ReorderBuffer *rb, ReorderBufferTXN *txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* ---------------------------------------
* toast reassembly support
* ---------------------------------------
*/
static void ReorderBufferToastInitHash(ReorderBuffer *rb, ReorderBufferTXN *txn);
static void ReorderBufferToastReset(ReorderBuffer *rb, ReorderBufferTXN *txn);
static void ReorderBufferToastReplace(ReorderBuffer *rb, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change);
static void ReorderBufferToastAppendChunk(ReorderBuffer *rb, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
* ---------------------------------------
* memory accounting
* ---------------------------------------
*/
static Size ReorderBufferChangeSize(ReorderBufferChange *change);
static void ReorderBufferChangeMemoryUpdate(ReorderBuffer *rb,
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
ReorderBufferChange *change,
bool addition, Size sz);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
* Allocate a new ReorderBuffer and clean out any old serialized state from
* prior ReorderBuffer instances for the same slot.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
ReorderBuffer *
ReorderBufferAllocate(void)
{
ReorderBuffer *buffer;
HASHCTL hash_ctl;
MemoryContext new_ctx;
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
Assert(MyReplicationSlot != NULL);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* allocate memory in own context, to have better accountability */
new_ctx = AllocSetContextCreate(CurrentMemoryContext,
"ReorderBuffer",
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);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
buffer =
(ReorderBuffer *) MemoryContextAlloc(new_ctx, sizeof(ReorderBuffer));
memset(&hash_ctl, 0, sizeof(hash_ctl));
buffer->context = new_ctx;
Use the new "Slab" context for some allocations in reorderbuffer.h. Note that this change alone does not yet fully address the performance problems triggering this work, a large portion of the slowdown is triggered by the tuple allocator, which isn't converted to the new allocator. It would be possible to do so, but using evenly sized objects, like both the current implementation in reorderbuffer.c and slab.c, wastes a fair amount of memory. A later patch by Tomas will introduce a better approach. Author: Tomas Vondra Reviewed-By: Andres Freund Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
2017-02-27 12:41:44 +01:00
buffer->change_context = SlabContextCreate(new_ctx,
"Change",
SLAB_DEFAULT_BLOCK_SIZE,
sizeof(ReorderBufferChange));
buffer->txn_context = SlabContextCreate(new_ctx,
"TXN",
SLAB_DEFAULT_BLOCK_SIZE,
sizeof(ReorderBufferTXN));
Improve the generation memory allocator Here we make a series of improvements to the generation memory allocator, namely: 1. Allow generation contexts to have a minimum, initial and maximum block sizes. The standard allocator allows this already but when the generation context was added, it only allowed fixed-sized blocks. The problem with fixed-sized blocks is that it's difficult to choose how large to make the blocks. If the chosen size is too small then we'd end up with a large number of blocks and a large number of malloc calls. If the block size is made too large, then memory is wasted. 2. Add support for "keeper" blocks. This is a special block that is allocated along with the context itself but is never freed. Instead, when the last chunk in the keeper block is freed, we simply mark the block as empty to allow new allocations to make use of it. 3. Add facility to "recycle" newly empty blocks instead of freeing them and having to later malloc an entire new block again. We do this by recording a single GenerationBlock which has become empty of any chunks. When we run out of space in the current block, we check to see if there is a "freeblock" and use that if it contains enough space for the allocation. Author: David Rowley, Tomas Vondra Reviewed-by: Andy Fan Discussion: https://postgr.es/m/d987fd54-01f8-0f73-af6c-519f799a0ab8@enterprisedb.com
2022-04-04 10:53:13 +02:00
/*
* XXX the allocation sizes used below pre-date generation context's block
* growing code. These values should likely be benchmarked and set to
* more suitable values.
*/
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
buffer->tup_context = GenerationContextCreate(new_ctx,
"Tuples",
Improve the generation memory allocator Here we make a series of improvements to the generation memory allocator, namely: 1. Allow generation contexts to have a minimum, initial and maximum block sizes. The standard allocator allows this already but when the generation context was added, it only allowed fixed-sized blocks. The problem with fixed-sized blocks is that it's difficult to choose how large to make the blocks. If the chosen size is too small then we'd end up with a large number of blocks and a large number of malloc calls. If the block size is made too large, then memory is wasted. 2. Add support for "keeper" blocks. This is a special block that is allocated along with the context itself but is never freed. Instead, when the last chunk in the keeper block is freed, we simply mark the block as empty to allow new allocations to make use of it. 3. Add facility to "recycle" newly empty blocks instead of freeing them and having to later malloc an entire new block again. We do this by recording a single GenerationBlock which has become empty of any chunks. When we run out of space in the current block, we check to see if there is a "freeblock" and use that if it contains enough space for the allocation. Author: David Rowley, Tomas Vondra Reviewed-by: Andy Fan Discussion: https://postgr.es/m/d987fd54-01f8-0f73-af6c-519f799a0ab8@enterprisedb.com
2022-04-04 10:53:13 +02:00
SLAB_LARGE_BLOCK_SIZE,
SLAB_LARGE_BLOCK_SIZE,
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
SLAB_LARGE_BLOCK_SIZE);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
hash_ctl.keysize = sizeof(TransactionId);
hash_ctl.entrysize = sizeof(ReorderBufferTXNByIdEnt);
hash_ctl.hcxt = buffer->context;
buffer->by_txn = hash_create("ReorderBufferByXid", 1000, &hash_ctl,
Improve hash_create's API for selecting simple-binary-key hash functions. Previously, if you wanted anything besides C-string hash keys, you had to specify a custom hashing function to hash_create(). Nearly all such callers were specifying tag_hash or oid_hash; which is tedious, and rather error-prone, since a caller could easily miss the opportunity to optimize by using hash_uint32 when appropriate. Replace this with a design whereby callers using simple binary-data keys just specify HASH_BLOBS and don't need to mess with specific support functions. hash_create() itself will take care of optimizing when the key size is four bytes. This nets out saving a few hundred bytes of code space, and offers a measurable performance improvement in tidbitmap.c (which was not exploiting the opportunity to use hash_uint32 for its 4-byte keys). There might be some wins elsewhere too, I didn't analyze closely. In future we could look into offering a similar optimized hashing function for 8-byte keys. Under this design that could be done in a centralized and machine-independent fashion, whereas getting it right for keys of platform-dependent sizes would've been notationally painful before. For the moment, the old way still works fine, so as not to break source code compatibility for loadable modules. Eventually we might want to remove tag_hash and friends from the exported API altogether, since there's no real need for them to be explicitly referenced from outside dynahash.c. Teodor Sigaev and Tom Lane
2014-12-18 19:36:29 +01:00
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
buffer->by_txn_last_xid = InvalidTransactionId;
buffer->by_txn_last_txn = NULL;
buffer->outbuf = NULL;
buffer->outbufsize = 0;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
buffer->size = 0;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
buffer->spillTxns = 0;
buffer->spillCount = 0;
buffer->spillBytes = 0;
Track statistics for streaming of changes from ReorderBuffer. This adds the statistics about transactions streamed to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. Commit 9868167500 has added the basic infrastructure to capture the stats of slot and this commit extends the statistics collector to track additional information about slots. Bump the catversion as we have added new columns in the catalog entry. Author: Ajin Cherian and Amit Kapila Reviewed-by: Sawada Masahiko and Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1+chpEomLzgSoky-D31qev19AmECNiEAietPQUGEFhtVA@mail.gmail.com
2020-10-29 04:41:51 +01:00
buffer->streamTxns = 0;
buffer->streamCount = 0;
buffer->streamBytes = 0;
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +02:00
buffer->totalTxns = 0;
buffer->totalBytes = 0;
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
buffer->current_restart_decoding_lsn = InvalidXLogRecPtr;
dlist_init(&buffer->toplevel_by_lsn);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
dlist_init(&buffer->txns_by_base_snapshot_lsn);
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
dclist_init(&buffer->catchange_txns);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
/*
* Ensure there's no stale data from prior uses of this slot, in case some
* prior exit avoided calling ReorderBufferFree. Failure to do this can
* produce duplicated txns, and it's very cheap if there's nothing there.
*/
ReorderBufferCleanupSerializedTXNs(NameStr(MyReplicationSlot->data.name));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return buffer;
}
/*
* Free a ReorderBuffer
*/
void
ReorderBufferFree(ReorderBuffer *rb)
{
MemoryContext context = rb->context;
/*
* We free separately allocated data by entirely scrapping reorderbuffer's
* memory context.
*/
MemoryContextDelete(context);
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
/* Free disk space used by unconsumed reorder buffers */
ReorderBufferCleanupSerializedTXNs(NameStr(MyReplicationSlot->data.name));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
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
* Get an unused, possibly preallocated, ReorderBufferTXN.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static ReorderBufferTXN *
ReorderBufferGetTXN(ReorderBuffer *rb)
{
ReorderBufferTXN *txn;
Use the new "Slab" context for some allocations in reorderbuffer.h. Note that this change alone does not yet fully address the performance problems triggering this work, a large portion of the slowdown is triggered by the tuple allocator, which isn't converted to the new allocator. It would be possible to do so, but using evenly sized objects, like both the current implementation in reorderbuffer.c and slab.c, wastes a fair amount of memory. A later patch by Tomas will introduce a better approach. Author: Tomas Vondra Reviewed-By: Andres Freund Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
2017-02-27 12:41:44 +01:00
txn = (ReorderBufferTXN *)
MemoryContextAlloc(rb->txn_context, sizeof(ReorderBufferTXN));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
memset(txn, 0, sizeof(ReorderBufferTXN));
dlist_init(&txn->changes);
dlist_init(&txn->tuplecids);
dlist_init(&txn->subtxns);
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
/* InvalidCommandId is not zero, so set it explicitly */
txn->command_id = InvalidCommandId;
Fix 'skip-empty-xacts' option in test_decoding for streaming mode. In streaming mode, the transaction can be decoded in multiple streams and those streams can be interleaved with streams of other transactions. So, we can't remember the transaction's write status in the logical decoding context because that might get changed due to some other transactions and lead to wrong answers for 'skip-empty-xacts' option. We decided to keep each transaction's write status in the ReorderBufferTxn to avoid interleaved streams changing the status of some unrelated transactions. Diagnosed-by: Amit Kapila Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/CAA4eK1LR7=XNM_TLmpZMFuV8ZQpoxkem--NZJYf8YXmesbvwLA@mail.gmail.com
2020-11-17 07:44:53 +01:00
txn->output_plugin_private = NULL;
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
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return txn;
}
/*
* Free a ReorderBufferTXN.
*/
Fix a bunch of functions that were declared static then defined not-static. Per testing with a compiler that whines about this.
2014-05-17 23:57:53 +02:00
static void
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferReturnTXN(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
/* clean the lookup cache if we were cached (quite likely) */
if (rb->by_txn_last_xid == txn->xid)
{
rb->by_txn_last_xid = InvalidTransactionId;
rb->by_txn_last_txn = NULL;
}
/* free data that's contained */
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
if (txn->gid != NULL)
{
pfree(txn->gid);
txn->gid = NULL;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (txn->tuplecid_hash != NULL)
{
hash_destroy(txn->tuplecid_hash);
txn->tuplecid_hash = NULL;
}
if (txn->invalidations)
{
pfree(txn->invalidations);
txn->invalidations = NULL;
}
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
/* Reset the toast hash */
ReorderBufferToastReset(rb, txn);
Use the new "Slab" context for some allocations in reorderbuffer.h. Note that this change alone does not yet fully address the performance problems triggering this work, a large portion of the slowdown is triggered by the tuple allocator, which isn't converted to the new allocator. It would be possible to do so, but using evenly sized objects, like both the current implementation in reorderbuffer.c and slab.c, wastes a fair amount of memory. A later patch by Tomas will introduce a better approach. Author: Tomas Vondra Reviewed-By: Andres Freund Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
2017-02-27 12:41:44 +01:00
pfree(txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
Fix some typos with {a,an} One of the changes impacts the documentation, so backpatch. Author: Peter Smith Discussion: https://postgr.es/m/CAHut+Pu6+c+r3mY24VT7u+H+E_s6vMr5OdRiZ8NT3EOa-E5Lmw@mail.gmail.com Backpatch-through: 14
2021-12-09 07:20:36 +01:00
* Get a fresh ReorderBufferChange.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
ReorderBufferChange *
ReorderBufferGetChange(ReorderBuffer *rb)
{
ReorderBufferChange *change;
Use the new "Slab" context for some allocations in reorderbuffer.h. Note that this change alone does not yet fully address the performance problems triggering this work, a large portion of the slowdown is triggered by the tuple allocator, which isn't converted to the new allocator. It would be possible to do so, but using evenly sized objects, like both the current implementation in reorderbuffer.c and slab.c, wastes a fair amount of memory. A later patch by Tomas will introduce a better approach. Author: Tomas Vondra Reviewed-By: Andres Freund Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
2017-02-27 12:41:44 +01:00
change = (ReorderBufferChange *)
MemoryContextAlloc(rb->change_context, sizeof(ReorderBufferChange));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
memset(change, 0, sizeof(ReorderBufferChange));
return change;
}
/*
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
* Free a ReorderBufferChange and update memory accounting, if requested.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
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
ReorderBufferReturnChange(ReorderBuffer *rb, ReorderBufferChange *change,
bool upd_mem)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* update memory accounting info */
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
if (upd_mem)
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
ReorderBufferChangeMemoryUpdate(rb, change, false,
ReorderBufferChangeSize(change));
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* free contained data */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
switch (change->action)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
case REORDER_BUFFER_CHANGE_INSERT:
case REORDER_BUFFER_CHANGE_UPDATE:
case REORDER_BUFFER_CHANGE_DELETE:
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT:
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (change->data.tp.newtuple)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferReturnTupleBuf(rb, change->data.tp.newtuple);
change->data.tp.newtuple = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (change->data.tp.oldtuple)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferReturnTupleBuf(rb, change->data.tp.oldtuple);
change->data.tp.oldtuple = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
break;
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
case REORDER_BUFFER_CHANGE_MESSAGE:
if (change->data.msg.prefix != NULL)
pfree(change->data.msg.prefix);
change->data.msg.prefix = NULL;
if (change->data.msg.message != NULL)
pfree(change->data.msg.message);
change->data.msg.message = NULL;
break;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
case REORDER_BUFFER_CHANGE_INVALIDATION:
if (change->data.inval.invalidations)
pfree(change->data.inval.invalidations);
change->data.inval.invalidations = NULL;
break;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT:
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (change->data.snapshot)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferFreeSnap(rb, change->data.snapshot);
change->data.snapshot = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
break;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/* no data in addition to the struct itself */
Fix memory leak in TRUNCATE decoding When decoding a TRUNCATE record, the relids array was being allocated in the main ReorderBuffer memory context, but not released with the change resulting in a memory leak. The array was also ignored when serializing/deserializing the change, assuming all the information is stored in the change itself. So when spilling the change to disk, we've only we have serialized only the pointer to the relids array. Thanks to never releasing the array, the pointer however remained valid even after loading the change back to memory, preventing an actual crash. This fixes both the memory leak and (de)serialization. The relids array is still allocated in the main ReorderBuffer memory context (none of the existing ones seems like a good match, and adding an extra context seems like an overkill). The allocation is wrapped in a new ReorderBuffer API functions, to keep the details within reorderbuffer.c, just like the other ReorderBufferGet methods do. Author: Tomas Vondra Discussion: https://www.postgresql.org/message-id/flat/66175a41-9342-2845-652f-1bd4c3ee50aa%402ndquadrant.com Backpatch: 11, where decoding of TRUNCATE was introduced
2018-09-03 02:10:24 +02:00
case REORDER_BUFFER_CHANGE_TRUNCATE:
if (change->data.truncate.relids != NULL)
{
ReorderBufferReturnRelids(rb, change->data.truncate.relids);
change->data.truncate.relids = NULL;
}
break;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM:
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT:
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID:
case REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID:
break;
}
Use the new "Slab" context for some allocations in reorderbuffer.h. Note that this change alone does not yet fully address the performance problems triggering this work, a large portion of the slowdown is triggered by the tuple allocator, which isn't converted to the new allocator. It would be possible to do so, but using evenly sized objects, like both the current implementation in reorderbuffer.c and slab.c, wastes a fair amount of memory. A later patch by Tomas will introduce a better approach. Author: Tomas Vondra Reviewed-By: Andres Freund Discussion: https://postgr.es/m/d15dff83-0b37-28ed-0809-95a5cc7292ad@2ndquadrant.com
2017-02-27 12:41:44 +01:00
pfree(change);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
Update obsolete comments Commit 9fab40ad32ef removed some pre-allocating logic in reorderbuffer.c, but left outdated comments in place. Repair. Author: Álvaro Herrera
2018-06-25 21:36:33 +02:00
* Get a fresh ReorderBufferTupleBuf fitting at least a tuple of size
* tuple_len (excluding header overhead).
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
ReorderBufferTupleBuf *
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
ReorderBufferGetTupleBuf(ReorderBuffer *rb, Size tuple_len)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
ReorderBufferTupleBuf *tuple;
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
Size alloc_len;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
alloc_len = tuple_len + SizeofHeapTupleHeader;
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
tuple = (ReorderBufferTupleBuf *)
MemoryContextAlloc(rb->tup_context,
sizeof(ReorderBufferTupleBuf) +
MAXIMUM_ALIGNOF + alloc_len);
tuple->alloc_tuple_size = alloc_len;
tuple->tuple.t_data = ReorderBufferTupleBufData(tuple);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return tuple;
}
/*
Fix some typos with {a,an} One of the changes impacts the documentation, so backpatch. Author: Peter Smith Discussion: https://postgr.es/m/CAHut+Pu6+c+r3mY24VT7u+H+E_s6vMr5OdRiZ8NT3EOa-E5Lmw@mail.gmail.com Backpatch-through: 14
2021-12-09 07:20:36 +01:00
* Free a ReorderBufferTupleBuf.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
ReorderBufferReturnTupleBuf(ReorderBuffer *rb, ReorderBufferTupleBuf *tuple)
{
Generational memory allocator Add new style of memory allocator, known as Generational appropriate for use in cases where memory is allocated and then freed in roughly oldest first order (FIFO). Use new allocator for logical decoding’s reorderbuffer to significantly reduce memory usage and improve performance. Author: Tomas Vondra Reviewed-by: Simon Riggs
2017-11-22 19:45:07 +01:00
pfree(tuple);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Fix memory leak in TRUNCATE decoding When decoding a TRUNCATE record, the relids array was being allocated in the main ReorderBuffer memory context, but not released with the change resulting in a memory leak. The array was also ignored when serializing/deserializing the change, assuming all the information is stored in the change itself. So when spilling the change to disk, we've only we have serialized only the pointer to the relids array. Thanks to never releasing the array, the pointer however remained valid even after loading the change back to memory, preventing an actual crash. This fixes both the memory leak and (de)serialization. The relids array is still allocated in the main ReorderBuffer memory context (none of the existing ones seems like a good match, and adding an extra context seems like an overkill). The allocation is wrapped in a new ReorderBuffer API functions, to keep the details within reorderbuffer.c, just like the other ReorderBufferGet methods do. Author: Tomas Vondra Discussion: https://www.postgresql.org/message-id/flat/66175a41-9342-2845-652f-1bd4c3ee50aa%402ndquadrant.com Backpatch: 11, where decoding of TRUNCATE was introduced
2018-09-03 02:10:24 +02:00
/*
* Get an array for relids of truncated relations.
*
* We use the global memory context (for the whole reorder buffer), because
* none of the existing ones seems like a good match (some are SLAB, so we
* can't use those, and tup_context is meant for tuple data, not relids). We
* could add yet another context, but it seems like an overkill - TRUNCATE is
* not particularly common operation, so it does not seem worth it.
*/
Oid *
ReorderBufferGetRelids(ReorderBuffer *rb, int nrelids)
{
Oid *relids;
Size alloc_len;
alloc_len = sizeof(Oid) * nrelids;
relids = (Oid *) MemoryContextAlloc(rb->context, alloc_len);
return relids;
}
/*
* Free an array of relids.
*/
void
ReorderBufferReturnRelids(ReorderBuffer *rb, Oid *relids)
{
pfree(relids);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Return the ReorderBufferTXN from the given buffer, specified by Xid.
* If create is true, and a transaction doesn't already exist, create it
* (with the given LSN, and as top transaction if that's specified);
* when this happens, is_new is set to true.
*/
static ReorderBufferTXN *
ReorderBufferTXNByXid(ReorderBuffer *rb, TransactionId xid, bool create,
bool *is_new, XLogRecPtr lsn, bool create_as_top)
{
ReorderBufferTXN *txn;
ReorderBufferTXNByIdEnt *ent;
bool found;
Assert(TransactionIdIsValid(xid));
/*
* Check the one-entry lookup cache first
*/
if (TransactionIdIsValid(rb->by_txn_last_xid) &&
rb->by_txn_last_xid == xid)
{
txn = rb->by_txn_last_txn;
if (txn != NULL)
{
/* found it, and it's valid */
if (is_new)
*is_new = false;
return txn;
}
/*
Fix various common mispellings. Mostly these are just comments but there are a few in documentation and a handful in code and tests. Hopefully this doesn't cause too much unnecessary pain for backpatching. I relented from some of the most common like "thru" for that reason. The rest don't seem numerous enough to cause problems. Thanks to Kevin Lyda's tool https://pypi.python.org/pypi/misspellings
2016-06-03 16:13:36 +02:00
* cached as non-existent, and asked not to create? Then nothing else
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* to do.
*/
if (!create)
return NULL;
/* otherwise fall through to create it */
}
/*
Fix some typos with {a,an} One of the changes impacts the documentation, so backpatch. Author: Peter Smith Discussion: https://postgr.es/m/CAHut+Pu6+c+r3mY24VT7u+H+E_s6vMr5OdRiZ8NT3EOa-E5Lmw@mail.gmail.com Backpatch-through: 14
2021-12-09 07:20:36 +01:00
* If the cache wasn't hit or it yielded a "does-not-exist" and we want to
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* create an entry.
*/
/* search the lookup table */
ent = (ReorderBufferTXNByIdEnt *)
hash_search(rb->by_txn,
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
&xid,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
create ? HASH_ENTER : HASH_FIND,
&found);
if (found)
txn = ent->txn;
else if (create)
{
/* initialize the new entry, if creation was requested */
Assert(ent != NULL);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
Assert(lsn != InvalidXLogRecPtr);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ent->txn = ReorderBufferGetTXN(rb);
ent->txn->xid = xid;
txn = ent->txn;
txn->first_lsn = lsn;
txn->restart_decoding_lsn = rb->current_restart_decoding_lsn;
if (create_as_top)
{
dlist_push_tail(&rb->toplevel_by_lsn, &txn->node);
AssertTXNLsnOrder(rb);
}
}
else
txn = NULL; /* not found and not asked to create */
/* update cache */
rb->by_txn_last_xid = xid;
rb->by_txn_last_txn = txn;
if (is_new)
*is_new = !found;
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(!create || txn != NULL);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return txn;
}
/*
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
* Record the partial change for the streaming of in-progress transactions. We
* can stream only complete changes so if we have a partial change like toast
* table insert or speculative insert then we mark such a 'txn' so that it
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
* can't be streamed. We also ensure that if the changes in such a 'txn' can
* be streamed and are above logical_decoding_work_mem threshold then we stream
* them as soon as we have a complete change.
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
*/
static void
ReorderBufferProcessPartialChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
ReorderBufferChange *change,
bool toast_insert)
{
ReorderBufferTXN *toptxn;
/*
* The partial changes need to be processed only while streaming
* in-progress transactions.
*/
if (!ReorderBufferCanStream(rb))
return;
/* Get the top transaction. */
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
toptxn = rbtxn_get_toptxn(txn);
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
/*
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
* Indicate a partial change for toast inserts. The change will be
* considered as complete once we get the insert or update on the main
* table and we are sure that the pending toast chunks are not required
* anymore.
*
* If we allow streaming when there are pending toast chunks then such
* chunks won't be released till the insert (multi_insert) is complete and
* we expect the txn to have streamed all changes after streaming. This
* restriction is mainly to ensure the correctness of streamed
* transactions and it doesn't seem worth uplifting such a restriction
* just to allow this case because anyway we will stream the transaction
* once such an insert is complete.
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
*/
if (toast_insert)
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
toptxn->txn_flags |= RBTXN_HAS_PARTIAL_CHANGE;
else if (rbtxn_has_partial_change(toptxn) &&
IsInsertOrUpdate(change->action) &&
change->data.tp.clear_toast_afterwards)
toptxn->txn_flags &= ~RBTXN_HAS_PARTIAL_CHANGE;
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
/*
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
* Indicate a partial change for speculative inserts. The change will be
Allow streaming the changes after speculative aborts. Until now, we didn't allow to stream the changes in logical replication till we receive speculative confirm or the next DML change record after speculative inserts. The reason was that we never use to process speculative aborts but after commit 4daa140a2f it is possible to process them so we can allow streaming once we receive speculative abort after speculative insertion. We decided to backpatch to 14 where the feature for streaming in progress transactions have been introduced as this is a minor change and makes that functionality better. Author: Amit Kapila Reviewed-By: Dilip Kumar Backpatch-through: 14 Discussion: https://postgr.es/m/CAA4eK1KdqmTCtrBR6oFfGELrLLbDLDedL6zACcsUOQuTJBj1vw@mail.gmail.com
2021-06-30 06:07:59 +02:00
* considered as complete once we get the speculative confirm or abort
* token.
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
*/
if (IsSpecInsert(change->action))
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
toptxn->txn_flags |= RBTXN_HAS_PARTIAL_CHANGE;
else if (rbtxn_has_partial_change(toptxn) &&
Allow streaming the changes after speculative aborts. Until now, we didn't allow to stream the changes in logical replication till we receive speculative confirm or the next DML change record after speculative inserts. The reason was that we never use to process speculative aborts but after commit 4daa140a2f it is possible to process them so we can allow streaming once we receive speculative abort after speculative insertion. We decided to backpatch to 14 where the feature for streaming in progress transactions have been introduced as this is a minor change and makes that functionality better. Author: Amit Kapila Reviewed-By: Dilip Kumar Backpatch-through: 14 Discussion: https://postgr.es/m/CAA4eK1KdqmTCtrBR6oFfGELrLLbDLDedL6zACcsUOQuTJBj1vw@mail.gmail.com
2021-06-30 06:07:59 +02:00
IsSpecConfirmOrAbort(change->action))
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
toptxn->txn_flags &= ~RBTXN_HAS_PARTIAL_CHANGE;
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
/*
* Stream the transaction if it is serialized before and the changes are
* now complete in the top-level transaction.
*
* The reason for doing the streaming of such a transaction as soon as we
* get the complete change for it is that previously it would have reached
* the memory threshold and wouldn't get streamed because of incomplete
* changes. Delaying such transactions would increase apply lag for them.
*/
if (ReorderBufferCanStartStreaming(rb) &&
Fix assertion during streaming of multi-insert toast changes. While decoding the multi-insert WAL we can't clean the toast untill we get the last insert of that WAL record. Now if we stream the changes before we get the last change, the memory for toast chunks won't be released and we expect the txn to have streamed all changes after streaming. This restriction is mainly to ensure the correctness of streamed transactions and it doesn't seem worth uplifting such a restriction just to allow this case because anyway we will stream the transaction once such an insert is complete. Previously we were using two different flags (one for toast tuples and another for speculative inserts) to indicate partial changes. Now instead we replaced both of them with a single flag to indicate partial changes. Reported-by: Pavan Deolasee Author: Dilip Kumar Reviewed-by: Pavan Deolasee, Amit Kapila Discussion: https://postgr.es/m/CABOikdN-_858zojYN-2tNcHiVTw-nhxPwoQS4quExeweQfG1Ug@mail.gmail.com
2021-05-27 04:29:43 +02:00
!(rbtxn_has_partial_change(toptxn)) &&
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
rbtxn_is_serialized(txn) &&
rbtxn_has_streamable_change(toptxn))
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
ReorderBufferStreamTXN(rb, toptxn);
}
/*
* Queue a change into a transaction so it can be replayed upon commit or will be
* streamed when we reach logical_decoding_work_mem threshold.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
ReorderBufferQueueChange(ReorderBuffer *rb, TransactionId xid, XLogRecPtr lsn,
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
ReorderBufferChange *change, bool toast_insert)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
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
/*
* While streaming the previous changes we have detected that the
* transaction is aborted. So there is no point in collecting further
* changes for it.
*/
if (txn->concurrent_abort)
{
/*
* We don't need to update memory accounting for this change as we
* have not added it to the queue yet.
*/
ReorderBufferReturnChange(rb, change, false);
return;
}
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
/*
* The changes that are sent downstream are considered streamable. We
* remember such transactions so that only those will later be considered
* for streaming.
*/
if (change->action == REORDER_BUFFER_CHANGE_INSERT ||
change->action == REORDER_BUFFER_CHANGE_UPDATE ||
change->action == REORDER_BUFFER_CHANGE_DELETE ||
change->action == REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT ||
change->action == REORDER_BUFFER_CHANGE_TRUNCATE ||
change->action == REORDER_BUFFER_CHANGE_MESSAGE)
{
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
ReorderBufferTXN *toptxn = rbtxn_get_toptxn(txn);
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
toptxn->txn_flags |= RBTXN_HAS_STREAMABLE_CHANGE;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
change->lsn = lsn;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
change->txn = txn;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(InvalidXLogRecPtr != lsn);
dlist_push_tail(&txn->changes, &change->node);
txn->nentries++;
txn->nentries_mem++;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* update memory accounting information */
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
ReorderBufferChangeMemoryUpdate(rb, change, true,
ReorderBufferChangeSize(change));
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
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
/* process partial change */
ReorderBufferProcessPartialChange(rb, txn, change, toast_insert);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* check the memory limits and evict something if needed */
ReorderBufferCheckMemoryLimit(rb);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
/*
Update comment atop of ReorderBufferQueueMessage(). The comments atop of this function describes behaviour in case of a transactional WAL message only, but it accepts both transactional and non-transactional WAL messages. Update the comments to describe behaviour in case of non-transactional WAL message as well. Ashutosh Bapat, rephrased by Amit Kapila Discussion: https://postgr.es/m/CAGEoWWTTzNzHOi8bj0wfAo1siGi-YEh6wqH1oaz4DrkTJ6HbTQ@mail.gmail.com
2020-12-19 05:38:46 +01:00
* A transactional message is queued to be processed upon commit and a
* non-transactional message gets processed immediately.
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
*/
void
ReorderBufferQueueMessage(ReorderBuffer *rb, TransactionId xid,
Harmonize reorderbuffer parameter names. Make reorderbuffer.h function declarations consistently use named parameters. Also make sure that the declarations use names that match corresponding names from function definitions in reorderbuffer.c. This makes the definitions easier to follow, especially in the case of functions that happen to have adjoining arguments of the same type. This patch was written with help from clang-tidy. Specifically, its "readability-inconsistent-declaration-parameter-name" check and its "readability-named-parameter" check were used. Author: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/3955318.1663377656@sss.pgh.pa.us
2022-09-18 02:20:17 +02:00
Snapshot snap, XLogRecPtr lsn,
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
bool transactional, const char *prefix,
Size message_size, const char *message)
{
if (transactional)
{
MemoryContext oldcontext;
ReorderBufferChange *change;
Assert(xid != InvalidTransactionId);
Fix snapshot handling in logicalmsg_decode Whe decoding a transactional logical message, logicalmsg_decode called SnapBuildGetOrBuildSnapshot. But we may not have a consistent snapshot yet at that point. We don't actually need the snapshot in this case (during replay we'll have the snapshot from the transaction), so in practice this is harmless. But in assert-enabled build this crashes. Fixed by requesting the snapshot only in non-transactional case, where we are guaranteed to have SNAPBUILD_CONSISTENT. Backpatch to 11. The issue exists since 9.6. Backpatch-through: 11 Reviewed-by: Andres Freund Discussion: https://postgr.es/m/84d60912-6eab-9b84-5de3-41765a5449e8@enterprisedb.com
2023-02-22 15:24:09 +01:00
/*
* We don't expect snapshots for transactional changes - we'll use the
* snapshot derived later during apply (unless the change gets
* skipped).
*/
Assert(!snap);
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
oldcontext = MemoryContextSwitchTo(rb->context);
change = ReorderBufferGetChange(rb);
change->action = REORDER_BUFFER_CHANGE_MESSAGE;
change->data.msg.prefix = pstrdup(prefix);
change->data.msg.message_size = message_size;
change->data.msg.message = palloc(message_size);
memcpy(change->data.msg.message, message, message_size);
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
ReorderBufferQueueChange(rb, xid, lsn, change, false);
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
MemoryContextSwitchTo(oldcontext);
}
else
{
ReorderBufferTXN *txn = NULL;
Harmonize reorderbuffer parameter names. Make reorderbuffer.h function declarations consistently use named parameters. Also make sure that the declarations use names that match corresponding names from function definitions in reorderbuffer.c. This makes the definitions easier to follow, especially in the case of functions that happen to have adjoining arguments of the same type. This patch was written with help from clang-tidy. Specifically, its "readability-inconsistent-declaration-parameter-name" check and its "readability-named-parameter" check were used. Author: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/3955318.1663377656@sss.pgh.pa.us
2022-09-18 02:20:17 +02:00
volatile Snapshot snapshot_now = snap;
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
Fix snapshot handling in logicalmsg_decode Whe decoding a transactional logical message, logicalmsg_decode called SnapBuildGetOrBuildSnapshot. But we may not have a consistent snapshot yet at that point. We don't actually need the snapshot in this case (during replay we'll have the snapshot from the transaction), so in practice this is harmless. But in assert-enabled build this crashes. Fixed by requesting the snapshot only in non-transactional case, where we are guaranteed to have SNAPBUILD_CONSISTENT. Backpatch to 11. The issue exists since 9.6. Backpatch-through: 11 Reviewed-by: Andres Freund Discussion: https://postgr.es/m/84d60912-6eab-9b84-5de3-41765a5449e8@enterprisedb.com
2023-02-22 15:24:09 +01:00
/* Non-transactional changes require a valid snapshot. */
Assert(snapshot_now);
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
if (xid != InvalidTransactionId)
txn = ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
/* setup snapshot to allow catalog access */
SetupHistoricSnapshot(snapshot_now, NULL);
PG_TRY();
{
rb->message(rb, txn, lsn, false, prefix, message_size, message);
TeardownHistoricSnapshot(false);
}
PG_CATCH();
{
TeardownHistoricSnapshot(true);
PG_RE_THROW();
}
PG_END_TRY();
}
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* AssertTXNLsnOrder
* Verify LSN ordering of transaction lists in the reorderbuffer
*
* Other LSN-related invariants are checked too.
*
* No-op if assertions are not in use.
*/
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void
AssertTXNLsnOrder(ReorderBuffer *rb)
{
#ifdef USE_ASSERT_CHECKING
Fix assertion failures while processing NEW_CID record in logical decoding. When the logical decoding restarts from NEW_CID, since there is no association between the top transaction and its subtransaction, both are created as top transactions and have the same LSN. This caused the assertion failure in AssertTXNLsnOrder(). This patch skips the assertion check until we reach the LSN at which we start decoding the contents of the transaction, specifically start_decoding_at LSN in SnapBuild. This is okay because we don't guarantee to make the association between top transaction and subtransaction until we try to decode the actual contents of transaction. The ordering of the records prior to the start_decoding_at LSN should have been checked before the restart. The other assertion failure is due to the reason that we forgot to track that we have considered top-level transaction id in the list of catalog changing transactions that were committed when one of its subtransactions is marked as containing catalog change. Reported-by: Tomas Vondra, Osumi Takamichi Author: Masahiko Sawada, Kuroda Hayato Reviewed-by: Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi, Masahiko Sawada Backpatch-through: 10 Discussion: https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.com Discussion: https://postgr.es/m/TYCPR01MB83733C6CEAE47D0280814D5AED7A9%40TYCPR01MB8373.jpnprd01.prod.outlook.com
2022-10-20 05:19:48 +02:00
LogicalDecodingContext *ctx = rb->private_data;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
dlist_iter iter;
XLogRecPtr prev_first_lsn = InvalidXLogRecPtr;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
XLogRecPtr prev_base_snap_lsn = InvalidXLogRecPtr;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Fix assertion failures while processing NEW_CID record in logical decoding. When the logical decoding restarts from NEW_CID, since there is no association between the top transaction and its subtransaction, both are created as top transactions and have the same LSN. This caused the assertion failure in AssertTXNLsnOrder(). This patch skips the assertion check until we reach the LSN at which we start decoding the contents of the transaction, specifically start_decoding_at LSN in SnapBuild. This is okay because we don't guarantee to make the association between top transaction and subtransaction until we try to decode the actual contents of transaction. The ordering of the records prior to the start_decoding_at LSN should have been checked before the restart. The other assertion failure is due to the reason that we forgot to track that we have considered top-level transaction id in the list of catalog changing transactions that were committed when one of its subtransactions is marked as containing catalog change. Reported-by: Tomas Vondra, Osumi Takamichi Author: Masahiko Sawada, Kuroda Hayato Reviewed-by: Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi, Masahiko Sawada Backpatch-through: 10 Discussion: https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.com Discussion: https://postgr.es/m/TYCPR01MB83733C6CEAE47D0280814D5AED7A9%40TYCPR01MB8373.jpnprd01.prod.outlook.com
2022-10-20 05:19:48 +02:00
/*
* Skip the verification if we don't reach the LSN at which we start
* decoding the contents of transactions yet because until we reach the
* LSN, we could have transactions that don't have the association between
* the top-level transaction and subtransaction yet and consequently have
* the same LSN. We don't guarantee this association until we try to
* decode the actual contents of transaction. The ordering of the records
* prior to the start_decoding_at LSN should have been checked before the
* restart.
*/
if (SnapBuildXactNeedsSkip(ctx->snapshot_builder, ctx->reader->EndRecPtr))
return;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
dlist_foreach(iter, &rb->toplevel_by_lsn)
{
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
ReorderBufferTXN *cur_txn = dlist_container(ReorderBufferTXN, node,
iter.cur);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* start LSN must be set */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(cur_txn->first_lsn != InvalidXLogRecPtr);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* If there is an end LSN, it must be higher than start LSN */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (cur_txn->end_lsn != InvalidXLogRecPtr)
Assert(cur_txn->first_lsn <= cur_txn->end_lsn);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* Current initial LSN must be strictly higher than previous */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (prev_first_lsn != InvalidXLogRecPtr)
Assert(prev_first_lsn < cur_txn->first_lsn);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* known-as-subtxn txns must not be listed */
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
Assert(!rbtxn_is_known_subxact(cur_txn));
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
prev_first_lsn = cur_txn->first_lsn;
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
dlist_foreach(iter, &rb->txns_by_base_snapshot_lsn)
{
ReorderBufferTXN *cur_txn = dlist_container(ReorderBufferTXN,
base_snapshot_node,
iter.cur);
/* base snapshot (and its LSN) must be set */
Assert(cur_txn->base_snapshot != NULL);
Assert(cur_txn->base_snapshot_lsn != InvalidXLogRecPtr);
/* current LSN must be strictly higher than previous */
if (prev_base_snap_lsn != InvalidXLogRecPtr)
Assert(prev_base_snap_lsn < cur_txn->base_snapshot_lsn);
/* known-as-subtxn txns must not be listed */
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
Assert(!rbtxn_is_known_subxact(cur_txn));
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
prev_base_snap_lsn = cur_txn->base_snapshot_lsn;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
#endif
}
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
/*
* AssertChangeLsnOrder
*
* Check ordering of changes in the (sub)transaction.
*/
static void
AssertChangeLsnOrder(ReorderBufferTXN *txn)
{
#ifdef USE_ASSERT_CHECKING
dlist_iter iter;
XLogRecPtr prev_lsn = txn->first_lsn;
dlist_foreach(iter, &txn->changes)
{
ReorderBufferChange *cur_change;
cur_change = dlist_container(ReorderBufferChange, node, iter.cur);
Assert(txn->first_lsn != InvalidXLogRecPtr);
Assert(cur_change->lsn != InvalidXLogRecPtr);
Assert(txn->first_lsn <= cur_change->lsn);
if (txn->end_lsn != InvalidXLogRecPtr)
Assert(cur_change->lsn <= txn->end_lsn);
Assert(prev_lsn <= cur_change->lsn);
prev_lsn = cur_change->lsn;
}
#endif
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* ReorderBufferGetOldestTXN
* Return oldest transaction in reorderbuffer
*/
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferTXN *
ReorderBufferGetOldestTXN(ReorderBuffer *rb)
{
ReorderBufferTXN *txn;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
AssertTXNLsnOrder(rb);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (dlist_is_empty(&rb->toplevel_by_lsn))
return NULL;
txn = dlist_head_element(ReorderBufferTXN, node, &rb->toplevel_by_lsn);
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
Assert(!rbtxn_is_known_subxact(txn));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(txn->first_lsn != InvalidXLogRecPtr);
return txn;
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* ReorderBufferGetOldestXmin
* Return oldest Xmin in reorderbuffer
*
* Returns oldest possibly running Xid from the point of view of snapshots
* used in the transactions kept by reorderbuffer, or InvalidTransactionId if
* there are none.
*
* Since snapshots are assigned monotonically, this equals the Xmin of the
* base snapshot with minimal base_snapshot_lsn.
*/
TransactionId
ReorderBufferGetOldestXmin(ReorderBuffer *rb)
{
ReorderBufferTXN *txn;
AssertTXNLsnOrder(rb);
if (dlist_is_empty(&rb->txns_by_base_snapshot_lsn))
return InvalidTransactionId;
txn = dlist_head_element(ReorderBufferTXN, base_snapshot_node,
&rb->txns_by_base_snapshot_lsn);
return txn->base_snapshot->xmin;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
void
ReorderBufferSetRestartPoint(ReorderBuffer *rb, XLogRecPtr ptr)
{
rb->current_restart_decoding_lsn = ptr;
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* ReorderBufferAssignChild
*
* Make note that we know that subxid is a subtransaction of xid, seen as of
* the given lsn.
*/
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
void
ReorderBufferAssignChild(ReorderBuffer *rb, TransactionId xid,
TransactionId subxid, XLogRecPtr lsn)
{
ReorderBufferTXN *txn;
ReorderBufferTXN *subtxn;
bool new_top;
bool new_sub;
txn = ReorderBufferTXNByXid(rb, xid, true, &new_top, lsn, true);
subtxn = ReorderBufferTXNByXid(rb, subxid, true, &new_sub, lsn, false);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
if (!new_sub)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_known_subxact(subtxn))
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
{
/* already associated, nothing to do */
return;
}
else
{
/*
* We already saw this transaction, but initially added it to the
* list of top-level txns. Now that we know it's not top-level,
* remove it from there.
*/
dlist_delete(&subtxn->node);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
subtxn->txn_flags |= RBTXN_IS_SUBXACT;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
subtxn->toplevel_xid = xid;
Assert(subtxn->nsubtxns == 0);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
/* set the reference to top-level transaction */
subtxn->toptxn = txn;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* add to subtransaction list */
dlist_push_tail(&txn->subtxns, &subtxn->node);
txn->nsubtxns++;
/* Possibly transfer the subtxn's snapshot to its top-level txn. */
ReorderBufferTransferSnapToParent(txn, subtxn);
/* Verify LSN-ordering invariant */
AssertTXNLsnOrder(rb);
}
/*
* ReorderBufferTransferSnapToParent
* Transfer base snapshot from subtxn to top-level txn, if needed
*
* This is done if the top-level txn doesn't have a base snapshot, or if the
* subtxn's base snapshot has an earlier LSN than the top-level txn's base
* snapshot's LSN. This can happen if there are no changes in the toplevel
* txn but there are some in the subtxn, or the first change in subtxn has
* earlier LSN than first change in the top-level txn and we learned about
* their kinship only now.
*
* The subtransaction's snapshot is cleared regardless of the transfer
* happening, since it's not needed anymore in either case.
*
* We do this as soon as we become aware of their kinship, to avoid queueing
* extra snapshots to txns known-as-subtxns -- only top-level txns will
* receive further snapshots.
*/
static void
ReorderBufferTransferSnapToParent(ReorderBufferTXN *txn,
ReorderBufferTXN *subtxn)
{
Assert(subtxn->toplevel_xid == txn->xid);
if (subtxn->base_snapshot != NULL)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
if (txn->base_snapshot == NULL ||
subtxn->base_snapshot_lsn < txn->base_snapshot_lsn)
{
/*
* If the toplevel transaction already has a base snapshot but
* it's newer than the subxact's, purge it.
*/
if (txn->base_snapshot != NULL)
{
SnapBuildSnapDecRefcount(txn->base_snapshot);
dlist_delete(&txn->base_snapshot_node);
}
/*
* The snapshot is now the top transaction's; transfer it, and
* adjust the list position of the top transaction in the list by
* moving it to where the subtransaction is.
*/
txn->base_snapshot = subtxn->base_snapshot;
txn->base_snapshot_lsn = subtxn->base_snapshot_lsn;
dlist_insert_before(&subtxn->base_snapshot_node,
&txn->base_snapshot_node);
/*
* The subtransaction doesn't have a snapshot anymore (so it
* mustn't be in the list.)
*/
subtxn->base_snapshot = NULL;
subtxn->base_snapshot_lsn = InvalidXLogRecPtr;
dlist_delete(&subtxn->base_snapshot_node);
}
else
{
/* Base snap of toplevel is fine, so subxact's is not needed */
SnapBuildSnapDecRefcount(subtxn->base_snapshot);
dlist_delete(&subtxn->base_snapshot_node);
subtxn->base_snapshot = NULL;
subtxn->base_snapshot_lsn = InvalidXLogRecPtr;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
}
/*
* Associate a subtransaction with its toplevel transaction at commit
* time. There may be no further changes added after this.
*/
void
ReorderBufferCommitChild(ReorderBuffer *rb, TransactionId xid,
TransactionId subxid, XLogRecPtr commit_lsn,
XLogRecPtr end_lsn)
{
ReorderBufferTXN *subtxn;
subtxn = ReorderBufferTXNByXid(rb, subxid, false, NULL,
InvalidXLogRecPtr, false);
/*
* No need to do anything if that subtxn didn't contain any changes
*/
if (!subtxn)
return;
subtxn->final_lsn = commit_lsn;
subtxn->end_lsn = end_lsn;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* Assign this subxact as a child of the toplevel xact (no-op if already
* done.)
*/
ReorderBufferAssignChild(rb, xid, subxid, InvalidXLogRecPtr);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Support for efficiently iterating over a transaction's and its
* subtransactions' changes.
*
* We do by doing a k-way merge between transactions/subtransactions. For that
* we model the current heads of the different transactions as a binary heap
* so we easily know which (sub-)transaction has the change with the smallest
* lsn next.
*
* We assume the changes in individual transactions are already sorted by LSN.
*/
/*
* Binary heap comparison function.
*/
static int
ReorderBufferIterCompare(Datum a, Datum b, void *arg)
{
ReorderBufferIterTXNState *state = (ReorderBufferIterTXNState *) arg;
XLogRecPtr pos_a = state->entries[DatumGetInt32(a)].lsn;
XLogRecPtr pos_b = state->entries[DatumGetInt32(b)].lsn;
if (pos_a < pos_b)
return 1;
else if (pos_a == pos_b)
return 0;
return -1;
}
/*
* Allocate & initialize an iterator which iterates in lsn order over a
* transaction and all its subtransactions.
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
*
* Note: The iterator state is returned through iter_state parameter rather
* than the function's return value. This is because the state gets cleaned up
* in a PG_CATCH block in the caller, so we want to make sure the caller gets
* back the state even if this function throws an exception.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
static void
ReorderBufferIterTXNInit(ReorderBuffer *rb, ReorderBufferTXN *txn,
ReorderBufferIterTXNState *volatile *iter_state)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Size nr_txns = 0;
ReorderBufferIterTXNState *state;
dlist_iter cur_txn_i;
int32 off;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
*iter_state = NULL;
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
/* Check ordering of changes in the toplevel transaction. */
AssertChangeLsnOrder(txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Calculate the size of our heap: one element for every transaction that
* contains changes. (Besides the transactions already in the reorder
* buffer, we count the one we were directly passed.)
*/
if (txn->nentries > 0)
nr_txns++;
dlist_foreach(cur_txn_i, &txn->subtxns)
{
ReorderBufferTXN *cur_txn;
cur_txn = dlist_container(ReorderBufferTXN, node, cur_txn_i.cur);
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
/* Check ordering of changes in this subtransaction. */
AssertChangeLsnOrder(cur_txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (cur_txn->nentries > 0)
nr_txns++;
}
/* allocate iteration state */
state = (ReorderBufferIterTXNState *)
MemoryContextAllocZero(rb->context,
sizeof(ReorderBufferIterTXNState) +
sizeof(ReorderBufferIterTXNEntry) * nr_txns);
state->nr_txns = nr_txns;
dlist_init(&state->old_change);
for (off = 0; off < state->nr_txns; off++)
{
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
state->entries[off].file.vfd = -1;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
state->entries[off].segno = 0;
}
/* allocate heap */
state->heap = binaryheap_allocate(state->nr_txns,
ReorderBufferIterCompare,
state);
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
/* Now that the state fields are initialized, it is safe to return it. */
*iter_state = state;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Now insert items into the binary heap, in an unordered fashion. (We
* will run a heap assembly step at the end; this is more efficient.)
*/
off = 0;
/* add toplevel transaction if it contains changes */
if (txn->nentries > 0)
{
ReorderBufferChange *cur_change;
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_serialized(txn))
Correct logical decoding restore behaviour for subtransactions. Before initializing iteration over a subtransaction's changes, the last few changes were not spilled to disk. That's correct if the transaction didn't spill to disk, but otherwise... This bug can lead to missed or misorderd subtransaction contents when they were spilled to disk. Move spilling of the remaining in-memory changes to ReorderBufferIterTXNInit(), where it can easily be applied to the top transaction and, if present, subtransactions. Since this code had too many bugs already, noticeably increase test coverage. Fixes: #14319 Reported-By: Huan Ruan Discussion: <20160909012610.20024.58169@wrigleys.postgresql.org> Backport: 9,4-, where logical decoding was added
2016-10-04 07:11:36 +02:00
{
/* serialize remaining changes */
ReorderBufferSerializeTXN(rb, txn);
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
ReorderBufferRestoreChanges(rb, txn, &state->entries[off].file,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
&state->entries[off].segno);
Correct logical decoding restore behaviour for subtransactions. Before initializing iteration over a subtransaction's changes, the last few changes were not spilled to disk. That's correct if the transaction didn't spill to disk, but otherwise... This bug can lead to missed or misorderd subtransaction contents when they were spilled to disk. Move spilling of the remaining in-memory changes to ReorderBufferIterTXNInit(), where it can easily be applied to the top transaction and, if present, subtransactions. Since this code had too many bugs already, noticeably increase test coverage. Fixes: #14319 Reported-By: Huan Ruan Discussion: <20160909012610.20024.58169@wrigleys.postgresql.org> Backport: 9,4-, where logical decoding was added
2016-10-04 07:11:36 +02:00
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
cur_change = dlist_head_element(ReorderBufferChange, node,
&txn->changes);
state->entries[off].lsn = cur_change->lsn;
state->entries[off].change = cur_change;
state->entries[off].txn = txn;
binaryheap_add_unordered(state->heap, Int32GetDatum(off++));
}
/* add subtransactions if they contain changes */
dlist_foreach(cur_txn_i, &txn->subtxns)
{
ReorderBufferTXN *cur_txn;
cur_txn = dlist_container(ReorderBufferTXN, node, cur_txn_i.cur);
if (cur_txn->nentries > 0)
{
ReorderBufferChange *cur_change;
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_serialized(cur_txn))
Correct logical decoding restore behaviour for subtransactions. Before initializing iteration over a subtransaction's changes, the last few changes were not spilled to disk. That's correct if the transaction didn't spill to disk, but otherwise... This bug can lead to missed or misorderd subtransaction contents when they were spilled to disk. Move spilling of the remaining in-memory changes to ReorderBufferIterTXNInit(), where it can easily be applied to the top transaction and, if present, subtransactions. Since this code had too many bugs already, noticeably increase test coverage. Fixes: #14319 Reported-By: Huan Ruan Discussion: <20160909012610.20024.58169@wrigleys.postgresql.org> Backport: 9,4-, where logical decoding was added
2016-10-04 07:11:36 +02:00
{
/* serialize remaining changes */
ReorderBufferSerializeTXN(rb, cur_txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferRestoreChanges(rb, cur_txn,
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
&state->entries[off].file,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
&state->entries[off].segno);
Correct logical decoding restore behaviour for subtransactions. Before initializing iteration over a subtransaction's changes, the last few changes were not spilled to disk. That's correct if the transaction didn't spill to disk, but otherwise... This bug can lead to missed or misorderd subtransaction contents when they were spilled to disk. Move spilling of the remaining in-memory changes to ReorderBufferIterTXNInit(), where it can easily be applied to the top transaction and, if present, subtransactions. Since this code had too many bugs already, noticeably increase test coverage. Fixes: #14319 Reported-By: Huan Ruan Discussion: <20160909012610.20024.58169@wrigleys.postgresql.org> Backport: 9,4-, where logical decoding was added
2016-10-04 07:11:36 +02:00
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
cur_change = dlist_head_element(ReorderBufferChange, node,
&cur_txn->changes);
state->entries[off].lsn = cur_change->lsn;
state->entries[off].change = cur_change;
state->entries[off].txn = cur_txn;
binaryheap_add_unordered(state->heap, Int32GetDatum(off++));
}
}
/* assemble a valid binary heap */
binaryheap_build(state->heap);
}
/*
* Return the next change when iterating over a transaction and its
* subtransactions.
*
* Returns NULL when no further changes exist.
*/
static ReorderBufferChange *
ReorderBufferIterTXNNext(ReorderBuffer *rb, ReorderBufferIterTXNState *state)
{
ReorderBufferChange *change;
ReorderBufferIterTXNEntry *entry;
int32 off;
/* nothing there anymore */
if (state->heap->bh_size == 0)
return NULL;
off = DatumGetInt32(binaryheap_first(state->heap));
entry = &state->entries[off];
/* free memory we might have "leaked" in the previous *Next call */
if (!dlist_is_empty(&state->old_change))
{
change = dlist_container(ReorderBufferChange, node,
dlist_pop_head_node(&state->old_change));
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(dlist_is_empty(&state->old_change));
}
change = entry->change;
/*
* update heap with information about which transaction has the next
* relevant change in LSN order
*/
/* there are in-memory changes */
if (dlist_has_next(&entry->txn->changes, &entry->change->node))
{
dlist_node *next = dlist_next_node(&entry->txn->changes, &change->node);
ReorderBufferChange *next_change =
dlist_container(ReorderBufferChange, node, next);
/* txn stays the same */
state->entries[off].lsn = next_change->lsn;
state->entries[off].change = next_change;
binaryheap_replace_first(state->heap, Int32GetDatum(off));
return change;
}
/* try to load changes from disk */
if (entry->txn->nentries != entry->txn->nentries_mem)
{
/*
* Ugly: restoring changes will reuse *Change records, thus delete the
* current one from the per-tx list and only free in the next call.
*/
dlist_delete(&change->node);
dlist_push_tail(&state->old_change, &change->node);
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +02:00
/*
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
* Update the total bytes processed by the txn for which we are
* releasing the current set of changes and restoring the new set of
* changes.
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +02:00
*/
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
rb->totalBytes += entry->txn->size;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
if (ReorderBufferRestoreChanges(rb, entry->txn, &entry->file,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
&state->entries[off].segno))
{
/* successfully restored changes from disk */
ReorderBufferChange *next_change =
dlist_head_element(ReorderBufferChange, node,
&entry->txn->changes);
elog(DEBUG2, "restored %u/%u changes from disk",
(uint32) entry->txn->nentries_mem,
(uint32) entry->txn->nentries);
Assert(entry->txn->nentries_mem);
/* txn stays the same */
state->entries[off].lsn = next_change->lsn;
state->entries[off].change = next_change;
binaryheap_replace_first(state->heap, Int32GetDatum(off));
return change;
}
}
/* ok, no changes there anymore, remove */
binaryheap_remove_first(state->heap);
return change;
}
/*
* Deallocate the iterator
*/
static void
ReorderBufferIterTXNFinish(ReorderBuffer *rb,
ReorderBufferIterTXNState *state)
{
int32 off;
for (off = 0; off < state->nr_txns; off++)
{
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
if (state->entries[off].file.vfd != -1)
FileClose(state->entries[off].file.vfd);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/* free memory we might have "leaked" in the last *Next call */
if (!dlist_is_empty(&state->old_change))
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node,
dlist_pop_head_node(&state->old_change));
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(dlist_is_empty(&state->old_change));
}
binaryheap_free(state->heap);
pfree(state);
}
/*
* Cleanup the contents of a transaction, usually after the transaction
* committed or aborted.
*/
static void
ReorderBufferCleanupTXN(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
bool found;
dlist_mutable_iter iter;
/* cleanup subtransactions & their changes */
dlist_foreach_modify(iter, &txn->subtxns)
{
ReorderBufferTXN *subtxn;
subtxn = dlist_container(ReorderBufferTXN, node, iter.cur);
/*
* Subtransactions are always associated to the toplevel TXN, even if
* they originally were happening inside another subtxn, so we won't
* ever recurse more than one level deep here.
*/
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
Assert(rbtxn_is_known_subxact(subtxn));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(subtxn->nsubtxns == 0);
ReorderBufferCleanupTXN(rb, subtxn);
}
Fix typos in logical.c and reorderbuffer.c. Reviewed-by: Sawada Masahiko Discussion: https://postgr.es/m/CAA4eK1K6zTpuqf_d7wXCBjo_EF0_B6Fz3Ecp71Vq18t=wG-nzg@mail.gmail.com
2020-10-09 04:45:53 +02:00
/* cleanup changes in the txn */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
dlist_foreach_modify(iter, &txn->changes)
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, iter.cur);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* Check we're not mixing changes from different transactions. */
Assert(change->txn == txn);
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Cleanup the tuplecids we stored for decoding catalog snapshot access.
* They are always stored in the toplevel transaction.
*/
dlist_foreach_modify(iter, &txn->tuplecids)
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, iter.cur);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* Check we're not mixing changes from different transactions. */
Assert(change->txn == txn);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(change->action == REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* Cleanup the base snapshot, if set.
*/
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (txn->base_snapshot != NULL)
{
SnapBuildSnapDecRefcount(txn->base_snapshot);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
dlist_delete(&txn->base_snapshot_node);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
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
/*
* Cleanup the snapshot for the last streamed run.
*/
if (txn->snapshot_now != NULL)
{
Assert(rbtxn_is_streamed(txn));
ReorderBufferFreeSnap(rb, txn->snapshot_now);
}
Remove duplicate code from ReorderBufferCleanupTXN(). Andres is apparently the only hacker who thinks this code is better as-is. I (tgl) follow some of his logic, but the fact that it's setting off warnings from static code analyzers seems like a sufficient reason to put the complexity into a comment rather than the code. Aleksander Alekseev Discussion: <20160404190345.54d84ee8@fujitsu>
2016-09-05 02:49:44 +02:00
/*
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
* Remove TXN from its containing lists.
Remove duplicate code from ReorderBufferCleanupTXN(). Andres is apparently the only hacker who thinks this code is better as-is. I (tgl) follow some of his logic, but the fact that it's setting off warnings from static code analyzers seems like a sufficient reason to put the complexity into a comment rather than the code. Aleksander Alekseev Discussion: <20160404190345.54d84ee8@fujitsu>
2016-09-05 02:49:44 +02:00
*
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
* Note: if txn is known as subxact, we are deleting the TXN from its
Remove duplicate code from ReorderBufferCleanupTXN(). Andres is apparently the only hacker who thinks this code is better as-is. I (tgl) follow some of his logic, but the fact that it's setting off warnings from static code analyzers seems like a sufficient reason to put the complexity into a comment rather than the code. Aleksander Alekseev Discussion: <20160404190345.54d84ee8@fujitsu>
2016-09-05 02:49:44 +02:00
* parent's list of known subxacts; this leaves the parent's nsubxacts
* count too high, but we don't care. Otherwise, we are deleting the TXN
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
* from the LSN-ordered list of toplevel TXNs. We remove the TXN from the
* list of catalog modifying transactions as well.
Remove duplicate code from ReorderBufferCleanupTXN(). Andres is apparently the only hacker who thinks this code is better as-is. I (tgl) follow some of his logic, but the fact that it's setting off warnings from static code analyzers seems like a sufficient reason to put the complexity into a comment rather than the code. Aleksander Alekseev Discussion: <20160404190345.54d84ee8@fujitsu>
2016-09-05 02:49:44 +02:00
*/
dlist_delete(&txn->node);
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
if (rbtxn_has_catalog_changes(txn))
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
dclist_delete_from(&rb->catchange_txns, &txn->catchange_node);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* now remove reference from buffer */
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(rb->by_txn, &txn->xid, HASH_REMOVE, &found);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(found);
/* remove entries spilled to disk */
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_serialized(txn))
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferRestoreCleanup(rb, txn);
/* deallocate */
ReorderBufferReturnTXN(rb, txn);
}
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
/*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* Discard changes from a transaction (and subtransactions), either after
* streaming or decoding them at PREPARE. Keep the remaining info -
* transactions, tuplecids, invalidations and snapshots.
*
Fix typos in pgstat.c, reorderbuffer.c and pathnodes.h Reviewed-by: Thomas Munro <thomas.munro@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Discussion: https://postgr.es/m/50250765-5B87-4AD7-9770-7FCED42A6175@yesql.se
2021-07-08 12:45:09 +02:00
* We additionally remove tuplecids after decoding the transaction at prepare
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* time as we only need to perform invalidation at rollback or commit prepared.
*
* 'txn_prepared' indicates that we have decoded the transaction at prepare
* time.
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
*/
static void
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
ReorderBufferTruncateTXN(ReorderBuffer *rb, ReorderBufferTXN *txn, bool txn_prepared)
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
{
dlist_mutable_iter iter;
/* cleanup subtransactions & their changes */
dlist_foreach_modify(iter, &txn->subtxns)
{
ReorderBufferTXN *subtxn;
subtxn = dlist_container(ReorderBufferTXN, node, iter.cur);
/*
* Subtransactions are always associated to the toplevel TXN, even if
* they originally were happening inside another subtxn, so we won't
* ever recurse more than one level deep here.
*/
Assert(rbtxn_is_known_subxact(subtxn));
Assert(subtxn->nsubtxns == 0);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
ReorderBufferTruncateTXN(rb, subtxn, txn_prepared);
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
}
Fix typos in logical.c and reorderbuffer.c. Reviewed-by: Sawada Masahiko Discussion: https://postgr.es/m/CAA4eK1K6zTpuqf_d7wXCBjo_EF0_B6Fz3Ecp71Vq18t=wG-nzg@mail.gmail.com
2020-10-09 04:45:53 +02:00
/* cleanup changes in the txn */
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
dlist_foreach_modify(iter, &txn->changes)
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, iter.cur);
/* Check we're not mixing changes from different transactions. */
Assert(change->txn == txn);
/* remove the change from it's containing list */
dlist_delete(&change->node);
ReorderBufferReturnChange(rb, change, true);
}
/*
* Mark the transaction as streamed.
*
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
* The top-level transaction, is marked as streamed always, even if it
* does not contain any changes (that is, when all the changes are in
* subtransactions).
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
*
* For subtransactions, we only mark them as streamed when there are
* changes in them.
*
* We do it this way because of aborts - we don't want to send aborts for
* XIDs the downstream is not aware of. And of course, it always knows
* about the toplevel xact (we send the XID in all messages), but we never
* stream XIDs of empty subxacts.
*/
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
if ((!txn_prepared) && (rbtxn_is_toptxn(txn) || (txn->nentries_mem != 0)))
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
txn->txn_flags |= RBTXN_IS_STREAMED;
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
if (txn_prepared)
{
/*
* If this is a prepared txn, cleanup the tuplecids we stored for
* decoding catalog snapshot access. They are always stored in the
* toplevel transaction.
*/
dlist_foreach_modify(iter, &txn->tuplecids)
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, iter.cur);
/* Check we're not mixing changes from different transactions. */
Assert(change->txn == txn);
Assert(change->action == REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID);
/* Remove the change from its containing list. */
dlist_delete(&change->node);
ReorderBufferReturnChange(rb, change, true);
}
}
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
/*
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
* Destroy the (relfilelocator, ctid) hashtable, so that we don't leak any
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
* memory. We could also keep the hash table and update it with new ctid
* values, but this seems simpler and good enough for now.
*/
if (txn->tuplecid_hash != NULL)
{
hash_destroy(txn->tuplecid_hash);
txn->tuplecid_hash = NULL;
}
/* If this txn is serialized then clean the disk space. */
if (rbtxn_is_serialized(txn))
{
ReorderBufferRestoreCleanup(rb, txn);
txn->txn_flags &= ~RBTXN_IS_SERIALIZED;
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
/*
* We set this flag to indicate if the transaction is ever serialized.
* We need this to accurately update the stats as otherwise the same
* transaction can be counted as serialized multiple times.
*/
txn->txn_flags |= RBTXN_IS_SERIALIZED_CLEAR;
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
}
/* also reset the number of entries in the transaction */
txn->nentries_mem = 0;
txn->nentries = 0;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +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
* Build a hash with a (relfilelocator, ctid) -> (cmin, cmax) mapping for use by
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
* HeapTupleSatisfiesHistoricMVCC.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static void
ReorderBufferBuildTupleCidHash(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
dlist_iter iter;
HASHCTL hash_ctl;
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (!rbtxn_has_catalog_changes(txn) || dlist_is_empty(&txn->tuplecids))
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return;
hash_ctl.keysize = sizeof(ReorderBufferTupleCidKey);
hash_ctl.entrysize = sizeof(ReorderBufferTupleCidEnt);
hash_ctl.hcxt = rb->context;
/*
* create the hash with the exact number of to-be-stored tuplecids from
* the start
*/
txn->tuplecid_hash =
hash_create("ReorderBufferTupleCid", txn->ntuplecids, &hash_ctl,
Improve hash_create's API for selecting simple-binary-key hash functions. Previously, if you wanted anything besides C-string hash keys, you had to specify a custom hashing function to hash_create(). Nearly all such callers were specifying tag_hash or oid_hash; which is tedious, and rather error-prone, since a caller could easily miss the opportunity to optimize by using hash_uint32 when appropriate. Replace this with a design whereby callers using simple binary-data keys just specify HASH_BLOBS and don't need to mess with specific support functions. hash_create() itself will take care of optimizing when the key size is four bytes. This nets out saving a few hundred bytes of code space, and offers a measurable performance improvement in tidbitmap.c (which was not exploiting the opportunity to use hash_uint32 for its 4-byte keys). There might be some wins elsewhere too, I didn't analyze closely. In future we could look into offering a similar optimized hashing function for 8-byte keys. Under this design that could be done in a centralized and machine-independent fashion, whereas getting it right for keys of platform-dependent sizes would've been notationally painful before. For the moment, the old way still works fine, so as not to break source code compatibility for loadable modules. Eventually we might want to remove tag_hash and friends from the exported API altogether, since there's no real need for them to be explicitly referenced from outside dynahash.c. Teodor Sigaev and Tom Lane
2014-12-18 19:36:29 +01:00
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
dlist_foreach(iter, &txn->tuplecids)
{
ReorderBufferTupleCidKey key;
ReorderBufferTupleCidEnt *ent;
bool found;
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, iter.cur);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(change->action == REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* be careful about padding */
memset(&key, 0, sizeof(ReorderBufferTupleCidKey));
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
key.rlocator = change->data.tuplecid.locator;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ItemPointerCopy(&change->data.tuplecid.tid,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
&key.tid);
ent = (ReorderBufferTupleCidEnt *)
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(txn->tuplecid_hash, &key, HASH_ENTER, &found);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (!found)
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ent->cmin = change->data.tuplecid.cmin;
ent->cmax = change->data.tuplecid.cmax;
ent->combocid = change->data.tuplecid.combocid;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
else
{
Relax overly strict assertion Ever since its birth, ReorderBufferBuildTupleCidHash() has contained an assertion that a catalog tuple cannot change Cmax after acquiring one. But that's wrong: if a subtransaction executes DDL that affects that catalog tuple, and later aborts and another DDL affects the same tuple, it will change Cmax. Relax the assertion to merely verify that the Cmax remains valid and monotonically increasing, instead. Add a test that tickles the relevant code. Diagnosed by, and initial patch submitted by: Arseny Sher Co-authored-by: Arseny Sher Discussion: https://postgr.es/m/874l9p8hyw.fsf@ars-thinkpad
2019-02-12 22:42:37 +01:00
/*
* Maybe we already saw this tuple before in this transaction, but
* if so it must have the same cmin.
*/
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(ent->cmin == change->data.tuplecid.cmin);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
Relax overly strict assertion Ever since its birth, ReorderBufferBuildTupleCidHash() has contained an assertion that a catalog tuple cannot change Cmax after acquiring one. But that's wrong: if a subtransaction executes DDL that affects that catalog tuple, and later aborts and another DDL affects the same tuple, it will change Cmax. Relax the assertion to merely verify that the Cmax remains valid and monotonically increasing, instead. Add a test that tickles the relevant code. Diagnosed by, and initial patch submitted by: Arseny Sher Co-authored-by: Arseny Sher Discussion: https://postgr.es/m/874l9p8hyw.fsf@ars-thinkpad
2019-02-12 22:42:37 +01:00
* cmax may be initially invalid, but once set it can only grow,
* and never become invalid again.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Relax overly strict assertion Ever since its birth, ReorderBufferBuildTupleCidHash() has contained an assertion that a catalog tuple cannot change Cmax after acquiring one. But that's wrong: if a subtransaction executes DDL that affects that catalog tuple, and later aborts and another DDL affects the same tuple, it will change Cmax. Relax the assertion to merely verify that the Cmax remains valid and monotonically increasing, instead. Add a test that tickles the relevant code. Diagnosed by, and initial patch submitted by: Arseny Sher Co-authored-by: Arseny Sher Discussion: https://postgr.es/m/874l9p8hyw.fsf@ars-thinkpad
2019-02-12 22:42:37 +01:00
Assert((ent->cmax == InvalidCommandId) ||
((change->data.tuplecid.cmax != InvalidCommandId) &&
(change->data.tuplecid.cmax > ent->cmax)));
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ent->cmax = change->data.tuplecid.cmax;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
}
}
/*
* Copy a provided snapshot so we can modify it privately. This is needed so
* that catalog modifying transactions can look into intermediate catalog
* states.
*/
static Snapshot
ReorderBufferCopySnap(ReorderBuffer *rb, Snapshot orig_snap,
ReorderBufferTXN *txn, CommandId cid)
{
Snapshot snap;
dlist_iter iter;
int i = 0;
Size size;
size = sizeof(SnapshotData) +
sizeof(TransactionId) * orig_snap->xcnt +
sizeof(TransactionId) * (txn->nsubtxns + 1);
snap = MemoryContextAllocZero(rb->context, size);
memcpy(snap, orig_snap, sizeof(SnapshotData));
snap->copied = true;
Fix assertion failure in logical decoding. Logical decoding set SnapshotData's regd_count field to avoid the snapshot manager from prematurely freeing snapshots that are generated by the decoding system. That was always an abuse of the field, as it was never supposed to be used outside the snapshot manager. Commit 94028691 made snapshot manager's tracking of the snapshots smarter, and that scheme fell apart. The snapshot manager got confused and hit the assertion, when a snapshot that was marked with regd_count==1 was not found in the heap, where the snapshot manager tracks registered the snapshots. To fix, don't abuse the regd_count field like that. Logical decoding still abuses the active_count field for similar purposes, but that's currently harmless. The assertion failure was first reported by Michael Paquier
2015-04-16 20:00:55 +02:00
snap->active_count = 1; /* mark as active so nobody frees it */
snap->regd_count = 0;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
snap->xip = (TransactionId *) (snap + 1);
memcpy(snap->xip, orig_snap->xip, sizeof(TransactionId) * snap->xcnt);
/*
* snap->subxip contains all txids that belong to our transaction which we
Add missing apostrophe. Masahiko Sawada Discussion: http://postgr.es/m/CAD21AoAzaR_XV7j7Wk9-QYXaFoT8H4egKwXvFY63wc8Lw2C9cg@mail.gmail.com
2017-05-15 21:41:15 +02:00
* need to check via cmin/cmax. That's why we store the toplevel
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* transaction in there as well.
*/
snap->subxip = snap->xip + snap->xcnt;
snap->subxip[i++] = txn->xid;
/*
Fix inconsistencies and typos in the tree, take 9 This addresses more issues with code comments, variable names and unreferenced variables. Author: Alexander Lakhin Discussion: https://postgr.es/m/7ab243e0-116d-3e44-d120-76b3df7abefd@gmail.com
2019-08-05 05:14:58 +02:00
* subxcnt isn't decreased when subtransactions abort, so count manually.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* Since it's an upper boundary it is safe to use it for the allocation
* above.
*/
snap->subxcnt = 1;
dlist_foreach(iter, &txn->subtxns)
{
ReorderBufferTXN *sub_txn;
sub_txn = dlist_container(ReorderBufferTXN, node, iter.cur);
snap->subxip[i++] = sub_txn->xid;
snap->subxcnt++;
}
/* sort so we can bsearch() later */
qsort(snap->subxip, snap->subxcnt, sizeof(TransactionId), xidComparator);
/* store the specified current CommandId */
snap->curcid = cid;
return snap;
}
/*
* Free a previously ReorderBufferCopySnap'ed snapshot
*/
static void
ReorderBufferFreeSnap(ReorderBuffer *rb, Snapshot snap)
{
if (snap->copied)
pfree(snap);
else
SnapBuildSnapDecRefcount(snap);
}
/*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* If the transaction was (partially) streamed, we need to prepare or commit
* it in a 'streamed' way. That is, we first stream the remaining part of the
* transaction, and then invoke stream_prepare or stream_commit message as per
* the case.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
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
static void
ReorderBufferStreamCommit(ReorderBuffer *rb, ReorderBufferTXN *txn)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
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
/* we should only call this for previously streamed transactions */
Assert(rbtxn_is_streamed(txn));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
ReorderBufferStreamTXN(rb, txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
if (rbtxn_prepared(txn))
{
/*
* Note, we send stream prepare even if a concurrent abort is
* detected. See DecodePrepare for more information.
*/
rb->stream_prepare(rb, txn, txn->final_lsn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* This is a PREPARED transaction, part of a two-phase commit. The
* full cleanup will happen as part of the COMMIT PREPAREDs, so now
Fix typos in comments The changes done in this commit impact comments with no direct user-visible changes, with fixes for incorrect function, variable or structure names. Author: Alexander Lakhin Discussion: https://postgr.es/m/e8c38840-596a-83d6-bd8d-cebc51111572@gmail.com
2023-05-02 05:23:08 +02:00
* just truncate txn by removing changes and tuplecids.
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
*/
ReorderBufferTruncateTXN(rb, txn, true);
/* Reset the CheckXidAlive */
CheckXidAlive = InvalidTransactionId;
}
else
{
rb->stream_commit(rb, txn, txn->final_lsn);
ReorderBufferCleanupTXN(rb, txn);
}
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
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
/*
* Set xid to detect concurrent aborts.
*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* While streaming an in-progress transaction or decoding a prepared
* transaction there is a possibility that the (sub)transaction might get
* aborted concurrently. In such case if the (sub)transaction has catalog
* update then we might decode the tuple using wrong catalog version. For
* example, suppose there is one catalog tuple with (xmin: 500, xmax: 0). Now,
* the transaction 501 updates the catalog tuple and after that we will have
* two tuples (xmin: 500, xmax: 501) and (xmin: 501, xmax: 0). Now, if 501 is
* aborted and some other transaction say 502 updates the same catalog tuple
* then the first tuple will be changed to (xmin: 500, xmax: 502). So, the
* problem is that when we try to decode the tuple inserted/updated in 501
* after the catalog update, we will see the catalog tuple with (xmin: 500,
* xmax: 502) as visible because it will consider that the tuple is deleted by
* xid 502 which is not visible to our snapshot. And when we will try to
* decode with that catalog tuple, it can lead to a wrong result or a crash.
* So, it is necessary to detect concurrent aborts to allow streaming of
Fix various typos and spelling mistakes in code comments Author: Justin Pryzby Discussion: https://postgr.es/m/20220411020336.GB26620@telsasoft.com
2022-04-11 10:49:41 +02:00
* in-progress transactions or decoding of prepared transactions.
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
*
* For detecting the concurrent abort we set CheckXidAlive to the current
* (sub)transaction's xid for which this change belongs to. And, during
* catalog scan we can check the status of the xid and if it is aborted we will
* report a specific error so that we can stop streaming current transaction
* and discard the already streamed changes on such an error. We might have
* already streamed some of the changes for the aborted (sub)transaction, but
* that is fine because when we decode the abort we will stream abort message
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* to truncate the changes in the subscriber. Similarly, for prepared
* transactions, we stop decoding if concurrent abort is detected and then
* rollback the changes when rollback prepared is encountered. See
Fix typo in reorderbuffer.c. Author: Peter Smith Discussion: https://postgr.es/m/CAHut+PtvzuYY0zu=dVRK_WVz5WGos1+otZWgEWqjha1ncoSRag@mail.gmail.com
2021-04-26 05:12:46 +02:00
* DecodePrepare.
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
*/
static inline void
SetupCheckXidLive(TransactionId xid)
{
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
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
* If the input transaction id is already set as a CheckXidAlive then
* nothing to do.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
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
if (TransactionIdEquals(CheckXidAlive, xid))
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return;
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
/*
* setup CheckXidAlive if it's not committed yet. We don't check if the
* xid is aborted. That will happen during catalog access.
*/
if (!TransactionIdDidCommit(xid))
CheckXidAlive = xid;
else
CheckXidAlive = InvalidTransactionId;
}
/*
* Helper function for ReorderBufferProcessTXN for applying change.
*/
static inline void
ReorderBufferApplyChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change,
bool streaming)
{
if (streaming)
rb->stream_change(rb, txn, relation, change);
else
rb->apply_change(rb, txn, relation, change);
}
/*
* Helper function for ReorderBufferProcessTXN for applying the truncate.
*/
static inline void
ReorderBufferApplyTruncate(ReorderBuffer *rb, ReorderBufferTXN *txn,
int nrelations, Relation *relations,
ReorderBufferChange *change, bool streaming)
{
if (streaming)
rb->stream_truncate(rb, txn, nrelations, relations, change);
else
rb->apply_truncate(rb, txn, nrelations, relations, change);
}
/*
* Helper function for ReorderBufferProcessTXN for applying the message.
*/
static inline void
ReorderBufferApplyMessage(ReorderBuffer *rb, ReorderBufferTXN *txn,
ReorderBufferChange *change, bool streaming)
{
if (streaming)
rb->stream_message(rb, txn, change->lsn, true,
change->data.msg.prefix,
change->data.msg.message_size,
change->data.msg.message);
else
rb->message(rb, txn, change->lsn, true,
change->data.msg.prefix,
change->data.msg.message_size,
change->data.msg.message);
}
/*
* Function to store the command id and snapshot at the end of the current
* stream so that we can reuse the same while sending the next stream.
*/
static inline void
ReorderBufferSaveTXNSnapshot(ReorderBuffer *rb, ReorderBufferTXN *txn,
Snapshot snapshot_now, CommandId command_id)
{
txn->command_id = command_id;
/* Avoid copying if it's already copied. */
if (snapshot_now->copied)
txn->snapshot_now = snapshot_now;
else
txn->snapshot_now = ReorderBufferCopySnap(rb, snapshot_now,
txn, command_id);
}
/*
* Helper function for ReorderBufferProcessTXN to handle the concurrent
* abort of the streaming transaction. This resets the TXN such that it
* can be used to stream the remaining data of transaction being processed.
Remove incorrect assertion in reorderbuffer.c. We start recording changes in ReorderBufferTXN even before we reach SNAPBUILD_CONSISTENT state so that if the commit is encountered after reaching that we should be able to send the changes of the entire transaction. Now, while recording changes if the reorder buffer memory has exceeded logical_decoding_work_mem then we can start streaming if it is allowed and we haven't yet streamed that data. However, we must not allow streaming to start unless the snapshot has reached SNAPBUILD_CONSISTENT state. In passing, improve the comments atop ReorderBufferResetTXN to mention the case when we need to continue streaming after getting an error. Author: Amit Kapila Reviewed-by: Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1KoOH0byboyYY40NBcC7Fe812trwTa+WY3jQF7WQWZbQg@mail.gmail.com
2020-12-04 09:24:50 +01:00
* This can happen when the subtransaction is aborted and we still want to
* continue processing the main or other subtransactions data.
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
*/
static void
ReorderBufferResetTXN(ReorderBuffer *rb, ReorderBufferTXN *txn,
Snapshot snapshot_now,
CommandId command_id,
XLogRecPtr last_lsn,
ReorderBufferChange *specinsert)
{
/* Discard the changes that we just streamed */
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
ReorderBufferTruncateTXN(rb, txn, rbtxn_prepared(txn));
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
/* Free all resources allocated for toast reconstruction */
ReorderBufferToastReset(rb, txn);
/* Return the spec insert change if it is not NULL */
if (specinsert != NULL)
{
ReorderBufferReturnChange(rb, specinsert, true);
specinsert = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* For the streaming case, stop the stream and remember the command ID and
* snapshot for the streaming run.
*/
if (rbtxn_is_streamed(txn))
{
rb->stream_stop(rb, txn, last_lsn);
ReorderBufferSaveTXNSnapshot(rb, txn, snapshot_now, command_id);
}
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
}
/*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* Helper function for ReorderBufferReplay and ReorderBufferStreamTXN.
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
*
* Send data of a transaction (and its subtransactions) to the
* output plugin. We iterate over the top and subtransactions (using a k-way
* merge) and replay the changes in lsn order.
*
* If streaming is true then data will be sent using stream API.
Remove useless Assert. Testing that an unsigned variable is >= 0 is pretty pointless, as noted by Coverity and numerous buildfarm members. In passing, add comment about new uses of "volatile" --- Coverity doesn't much like that either, but it seems probably necessary.
2020-08-09 17:32:31 +02:00
*
* Note: "volatile" markers on some parameters are to avoid trouble with
* PG_TRY inside the function.
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
*/
static void
ReorderBufferProcessTXN(ReorderBuffer *rb, ReorderBufferTXN *txn,
XLogRecPtr commit_lsn,
volatile Snapshot snapshot_now,
volatile CommandId command_id,
bool streaming)
{
bool using_subtxn;
MemoryContext ccxt = CurrentMemoryContext;
ReorderBufferIterTXNState *volatile iterstate = NULL;
volatile XLogRecPtr prev_lsn = InvalidXLogRecPtr;
ReorderBufferChange *volatile specinsert = NULL;
volatile bool stream_started = false;
ReorderBufferTXN *volatile curtxn = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* build data to be able to lookup the CommandIds of catalog tuples */
ReorderBufferBuildTupleCidHash(rb, txn);
/* setup the initial snapshot */
SetupHistoricSnapshot(snapshot_now, txn->tuplecid_hash);
Further cleanup of ReorderBufferCommit(). On closer inspection, we can remove the "volatile" qualifier on "using_subtxn" so long as we initialize that before the PG_TRY block, which there's no particularly good reason not to do. Also, push the "change" variable inside the PG_TRY so as to remove all question of whether it needs "volatile", and remove useless early initializations of "snapshow_now" and "using_subtxn".
2015-01-26 04:49:56 +01:00
/*
* Decoding needs access to syscaches et al., which in turn use
* heavyweight locks and such. Thus we need to have enough state around to
* keep track of those. The easiest way is to simply use a transaction
* internally. That also allows us to easily enforce that nothing writes
* to the database by checking for xid assignments.
*
* When we're called via the SQL SRF there's already a transaction
* started, so start an explicit subtransaction there.
*/
using_subtxn = IsTransactionOrTransactionBlock();
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
PG_TRY();
{
Further cleanup of ReorderBufferCommit(). On closer inspection, we can remove the "volatile" qualifier on "using_subtxn" so long as we initialize that before the PG_TRY block, which there's no particularly good reason not to do. Also, push the "change" variable inside the PG_TRY so as to remove all question of whether it needs "volatile", and remove useless early initializations of "snapshow_now" and "using_subtxn".
2015-01-26 04:49:56 +01:00
ReorderBufferChange *change;
Fix the logical replication timeout during large DDLs. The DDLs like Refresh Materialized views that generate lots of temporary data due to rewrite rules may not be processed by output plugins (for example pgoutput). So, we won't send keep-alive messages for a long time while processing such commands and that can lead the subscriber side to timeout. We have previously fixed a similar case for large transactions in commit f95d53eded where the output plugin filters all or most of the changes but missed to handle the DDLs. We decided not to backpatch this as this adds a new callback in the existing exposed structure and moreover, users can increase the wal_sender_timeout and wal_receiver_timeout to avoid this problem. Author: Wang wei, Hou Zhijie Reviewed-by: Peter Smith, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/OS3PR01MB6275478E5D29E4A563302D3D9E2B9@OS3PR01MB6275.jpnprd01.prod.outlook.com Discussion: https://postgr.es/m/CAA5-nLARN7-3SLU_QUxfy510pmrYK6JJb=bk3hcgemAM_pAv+w@mail.gmail.com
2023-02-08 03:28:25 +01:00
int changes_count = 0; /* used to accumulate the number of
* changes */
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
if (using_subtxn)
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
BeginInternalSubTransaction(streaming ? "stream" : "replay");
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
else
StartTransactionCommand();
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* We only need to send begin/begin-prepare for non-streamed
* transactions.
*/
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
if (!streaming)
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
{
if (rbtxn_prepared(txn))
rb->begin_prepare(rb, txn);
else
rb->begin(rb, txn);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
ReorderBufferIterTXNInit(rb, txn, &iterstate);
Further cleanup of ReorderBufferCommit(). On closer inspection, we can remove the "volatile" qualifier on "using_subtxn" so long as we initialize that before the PG_TRY block, which there's no particularly good reason not to do. Also, push the "change" variable inside the PG_TRY so as to remove all question of whether it needs "volatile", and remove useless early initializations of "snapshow_now" and "using_subtxn".
2015-01-26 04:49:56 +01:00
while ((change = ReorderBufferIterTXNNext(rb, iterstate)) != NULL)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Relation relation = NULL;
Oid reloid;
Add CHECK_FOR_INTERRUPTS while decoding changes. While decoding changes in a loop, if we skip all the changes there is no CFI making the loop uninterruptible. Reported-by: Whale Song and Andrey Borodin Bug: 17580 Author: Masahiko Sawada Reviwed-by: Amit Kapila Backpatch-through: 10 Discussion: https://postgr.es/m/17580-849c1d5b6d7eb422@postgresql.org Discussion: https://postgr.es/m/B319ECD6-9A28-4CDF-A8F4-3591E0BF2369@yandex-team.ru
2022-08-23 06:50:02 +02:00
CHECK_FOR_INTERRUPTS();
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
/*
* We can't call start stream callback before processing first
* change.
*/
if (prev_lsn == InvalidXLogRecPtr)
{
if (streaming)
{
txn->origin_id = change->origin_id;
rb->stream_start(rb, txn, change->lsn);
stream_started = true;
}
}
/*
* Enforce correct ordering of changes, merged from multiple
* subtransactions. The changes may have the same LSN due to
* MULTI_INSERT xlog records.
*/
Assert(prev_lsn == InvalidXLogRecPtr || prev_lsn <= change->lsn);
prev_lsn = change->lsn;
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* Set the current xid to detect concurrent aborts. This is
* required for the cases when we decode the changes before the
* COMMIT record is processed.
*/
if (streaming || rbtxn_prepared(change->txn))
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
{
curtxn = change->txn;
SetupCheckXidLive(curtxn->xid);
}
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
switch (change->action)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM:
pgindent run for 9.5
2015-05-24 03:35:49 +02:00
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/*
* Confirmation for speculative insertion arrived. Simply
* use as a normal record. It'll be cleaned up at the end
* of INSERT processing.
*/
logical decoding: beware of an unset specinsert change Coverity complains that there is no protection in the code (at least in non-assertion-enabled builds) against speculative insertion failing to follow the expected protocol. Add an elog(ERROR) for the case.
2018-07-05 23:42:37 +02:00
if (specinsert == NULL)
elog(ERROR, "invalid ordering of speculative insertion changes");
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
Assert(specinsert->data.tp.oldtuple == NULL);
change = specinsert;
change->action = REORDER_BUFFER_CHANGE_INSERT;
Dial back -Wimplicit-fallthrough to level 3 The additional pain from level 4 is excessive for the gain. Also revert all the source annotation changes to their original wordings, to avoid back-patching pain. Discussion: https://postgr.es/m/31166.1589378554@sss.pgh.pa.us
2020-05-13 21:31:14 +02:00
/* intentionally fall through */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
case REORDER_BUFFER_CHANGE_INSERT:
case REORDER_BUFFER_CHANGE_UPDATE:
case REORDER_BUFFER_CHANGE_DELETE:
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(snapshot_now);
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
reloid = RelidByRelfilenumber(change->data.tp.rlocator.spcOid,
change->data.tp.rlocator.relNumber);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
Fix logical decoding error when system table w/ toast is repeatedly rewritten. Repeatedly rewriting a mapped catalog table with VACUUM FULL or CLUSTER could cause logical decoding to fail with: ERROR, "could not map filenode \"%s\" to relation OID" To trigger the problem the rewritten catalog had to have live tuples with toasted columns. The problem was triggered as during catalog table rewrites the heap_insert() check that prevents logical decoding information to be emitted for system catalogs, failed to treat the new heap's toast table as a system catalog (because the new heap is not recognized as a catalog table via RelationIsLogicallyLogged()). The relmapper, in contrast to the normal catalog contents, does not contain historical information. After a single rewrite of a mapped table the new relation is known to the relmapper, but if the table is rewritten twice before logical decoding occurs, the relfilenode cannot be mapped to a relation anymore. Which then leads us to error out. This only happens for toast tables, because the main table contents aren't re-inserted with heap_insert(). The fix is simple, add a new heap_insert() flag that prevents logical decoding information from being emitted, and accept during decoding that there might not be tuple data for toast tables. Unfortunately that does not fix pre-existing logical decoding errors. Doing so would require not throwing an error when a filenode cannot be mapped to a relation during decoding, and that seems too likely to hide bugs. If it's crucial to fix decoding for an existing slot, temporarily changing the ERROR in ReorderBufferCommit() to a WARNING appears to be the best fix. Author: Andres Freund Discussion: https://postgr.es/m/20180914021046.oi7dm4ra3ot2g2kt@alap3.anarazel.de Backpatch: 9.4-, where logical decoding was introduced
2018-10-10 22:53:02 +02:00
* Mapped catalog tuple without data, emitted while
* catalog table was in the process of being rewritten. We
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
* can fail to look up the relfilenumber, because the
Fix typo in reorderbuffer.c. Author: Zhijie Hou Reviewed-by: Sawada Masahiko Discussion: https://postgr.es/m/ba88bb58aaf14284abca16aec04bf279@G08CNEXMBPEKD05.g08.fujitsu.local
2021-01-05 03:26:40 +01:00
* relmapper has no "historic" view, in contrast to the
* normal catalog during decoding. Thus repeated rewrites
* can cause a lookup failure. That's OK because we do not
* decode catalog changes anyway. Normally such tuples
* would be skipped over below, but we can't identify
* whether the table should be logically logged without
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
* mapping the relfilenumber to the oid.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
if (reloid == InvalidOid &&
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
change->data.tp.newtuple == NULL &&
change->data.tp.oldtuple == NULL)
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
goto change_done;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
else if (reloid == InvalidOid)
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
elog(ERROR, "could not map filenumber \"%s\" to relation OID",
relpathperm(change->data.tp.rlocator,
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
MAIN_FORKNUM));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
relation = RelationIdGetRelation(reloid);
Fix RelationIdGetRelation calls that weren't bothering with error checks. Some of these are quite old, but that doesn't make them not bugs. We'd rather report a failure via elog than SIGSEGV. While at it, uniformly spell the error check as !RelationIsValid(rel) rather than a bare rel == NULL test. The machine code is the same but it seems better to be consistent. Coverity complained about this today, not sure why, because the mistake is in fact old.
2019-09-08 23:00:29 +02:00
if (!RelationIsValid(relation))
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
elog(ERROR, "could not open relation with OID %u (for filenumber \"%s\")",
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
reloid,
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
relpathperm(change->data.tp.rlocator,
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
MAIN_FORKNUM));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
if (!RelationIsLogicallyLogged(relation))
goto change_done;
Handle heap rewrites even better in logical decoding Logical decoding should not publish anything about tables created as part of a heap rewrite during DDL. Those tables don't exist externally, so consumers of logical decoding cannot do anything sensible with that information. In ab28feae2bd3d4629bd73ae3548e671c57d785f0, we worked around this for built-in logical replication, but that was hack. This is a more proper fix: We mark such transient heaps using the new field pg_class.relwrite, linking to the original relation OID. By default, we ignore them in logical decoding before they get to the output plugin. Optionally, a plugin can register their interest in getting such changes, if they handle DDL specially, in which case the new field will help them get information about the actual table. Reviewed-by: Craig Ringer <craig@2ndquadrant.com>
2018-03-21 14:13:24 +01:00
/*
* Ignore temporary heaps created during DDL unless the
* plugin has asked for them.
*/
if (relation->rd_rel->relrewrite && !rb->output_rewrites)
goto change_done;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/*
* For now ignore sequence changes entirely. Most of the
* time they don't log changes using records we
* understand, so it doesn't make sense to handle the few
* cases we do.
*/
if (relation->rd_rel->relkind == RELKIND_SEQUENCE)
goto change_done;
/* user-triggered change */
if (!IsToastRelation(relation))
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
ReorderBufferToastReplace(rb, txn, relation, change);
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
ReorderBufferApplyChange(rb, txn, relation, change,
streaming);
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
* Only clear reassembled toast chunks if we're sure
* they're not required anymore. The creator of the
* tuple tells us.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
if (change->data.tp.clear_toast_afterwards)
ReorderBufferToastReset(rb, txn);
}
/* we're not interested in toast deletions */
else if (change->action == REORDER_BUFFER_CHANGE_INSERT)
{
/*
* Need to reassemble the full toasted Datum in
* memory, to ensure the chunks don't get reused till
* we're done remove it from the list of this
* transaction's changes. Otherwise it will get
* freed/reused while restoring spooled data from
* disk.
*/
Do not decode TOAST data for table rewrites During table rewrites (VACUUM FULL and CLUSTER), the main heap is logged using XLOG / FPI records, and thus (correctly) ignored in decoding. But the associated TOAST table is WAL-logged as plain INSERT records, and so was logically decoded and passed to reorder buffer. That has severe consequences with TOAST tables of non-trivial size. Firstly, reorder buffer has to keep all those changes, possibly spilling them to a file, incurring I/O costs and disk space. Secondly, ReoderBufferCommit() was stashing all those TOAST chunks into a hash table, which got discarded only after processing the row from the main heap. But as the main heap is not decoded for rewrites, this never happened, so all the TOAST data accumulated in memory, resulting either in excessive memory consumption or OOM. The fix is simple, as commit e9edc1ba already introduced infrastructure (namely HEAP_INSERT_NO_LOGICAL flag) to skip logical decoding of TOAST tables, but it only applied it to system tables. So simply use it for all TOAST data in raw_heap_insert(). That would however solve only the memory consumption issue - the TOAST changes would still be decoded and added to the reorder buffer, and spilled to disk (although without TOAST tuple data, so much smaller). But we can solve that by tweaking DecodeInsert() to just ignore such INSERT records altogether, using XLH_INSERT_CONTAINS_NEW_TUPLE flag, instead of skipping them later in ReorderBufferCommit(). Review: Masahiko Sawada Discussion: https://www.postgresql.org/message-id/flat/1a17c643-e9af-3dba-486b-fbe31bc1823a%402ndquadrant.com Backpatch: 9.4-, where logical decoding was introduced
2018-11-28 01:11:15 +01:00
Assert(change->data.tp.newtuple != NULL);
dlist_delete(&change->node);
ReorderBufferToastAppendChunk(rb, txn, relation,
change);
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
}
change_done:
pgindent run for 9.5
2015-05-24 03:35:49 +02:00
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/*
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
* If speculative insertion was confirmed, the record
* isn't needed anymore.
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
*/
if (specinsert != NULL)
{
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
ReorderBufferReturnChange(rb, specinsert, true);
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
specinsert = NULL;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
if (RelationIsValid(relation))
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
{
RelationClose(relation);
relation = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
break;
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT:
pgindent run for 9.5
2015-05-24 03:35:49 +02:00
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/*
* Speculative insertions are dealt with by delaying the
* processing of the insert until the confirmation record
* arrives. For that we simply unlink the record from the
* chain, so it does not get freed/reused while restoring
* spooled data from disk.
*
* This is safe in the face of concurrent catalog changes
* because the relevant relation can't be changed between
* speculative insertion and confirmation due to
* CheckTableNotInUse() and locking.
*/
/* clear out a pending (and thus failed) speculation */
if (specinsert != NULL)
{
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
ReorderBufferReturnChange(rb, specinsert, true);
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
specinsert = NULL;
}
/* and memorize the pending insertion */
dlist_delete(&change->node);
specinsert = change;
break;
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT:
/*
* Abort for speculative insertion arrived. So cleanup the
* specinsert tuple and toast hash.
*
* Note that we get the spec abort change for each toast
* entry but we need to perform the cleanup only the first
* time we get it for the main table.
*/
if (specinsert != NULL)
{
/*
* We must clean the toast hash before processing a
* completely new tuple to avoid confusion about the
* previous tuple's toast chunks.
*/
Assert(change->data.tp.clear_toast_afterwards);
ReorderBufferToastReset(rb, txn);
/* We don't need this record anymore. */
ReorderBufferReturnChange(rb, specinsert, true);
specinsert = NULL;
}
break;
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
case REORDER_BUFFER_CHANGE_TRUNCATE:
{
Post-feature-freeze pgindent run. Discussion: https://postgr.es/m/15719.1523984266@sss.pgh.pa.us
2018-04-26 20:47:16 +02:00
int i;
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
int nrelids = change->data.truncate.nrelids;
int nrelations = 0;
Relation *relations;
relations = palloc0(nrelids * sizeof(Relation));
for (i = 0; i < nrelids; i++)
Post-feature-freeze pgindent run. Discussion: https://postgr.es/m/15719.1523984266@sss.pgh.pa.us
2018-04-26 20:47:16 +02:00
{
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
Oid relid = change->data.truncate.relids[i];
Rename shadowed local variables In a similar effort to f01592f91, here we mostly rename shadowed local variables to remove the warnings produced when compiling with -Wshadow=compatible-local. This fixes 63 warnings and leaves just 5. Author: Justin Pryzby, David Rowley Reviewed-by: Justin Pryzby Discussion https://postgr.es/m/20220817145434.GC26426%40telsasoft.com
2022-10-05 10:01:41 +02:00
Relation rel;
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
Rename shadowed local variables In a similar effort to f01592f91, here we mostly rename shadowed local variables to remove the warnings produced when compiling with -Wshadow=compatible-local. This fixes 63 warnings and leaves just 5. Author: Justin Pryzby, David Rowley Reviewed-by: Justin Pryzby Discussion https://postgr.es/m/20220817145434.GC26426%40telsasoft.com
2022-10-05 10:01:41 +02:00
rel = RelationIdGetRelation(relid);
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
Rename shadowed local variables In a similar effort to f01592f91, here we mostly rename shadowed local variables to remove the warnings produced when compiling with -Wshadow=compatible-local. This fixes 63 warnings and leaves just 5. Author: Justin Pryzby, David Rowley Reviewed-by: Justin Pryzby Discussion https://postgr.es/m/20220817145434.GC26426%40telsasoft.com
2022-10-05 10:01:41 +02:00
if (!RelationIsValid(rel))
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
elog(ERROR, "could not open relation with OID %u", relid);
Rename shadowed local variables In a similar effort to f01592f91, here we mostly rename shadowed local variables to remove the warnings produced when compiling with -Wshadow=compatible-local. This fixes 63 warnings and leaves just 5. Author: Justin Pryzby, David Rowley Reviewed-by: Justin Pryzby Discussion https://postgr.es/m/20220817145434.GC26426%40telsasoft.com
2022-10-05 10:01:41 +02:00
if (!RelationIsLogicallyLogged(rel))
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
continue;
Rename shadowed local variables In a similar effort to f01592f91, here we mostly rename shadowed local variables to remove the warnings produced when compiling with -Wshadow=compatible-local. This fixes 63 warnings and leaves just 5. Author: Justin Pryzby, David Rowley Reviewed-by: Justin Pryzby Discussion https://postgr.es/m/20220817145434.GC26426%40telsasoft.com
2022-10-05 10:01:41 +02:00
relations[nrelations++] = rel;
Post-feature-freeze pgindent run. Discussion: https://postgr.es/m/15719.1523984266@sss.pgh.pa.us
2018-04-26 20:47:16 +02:00
}
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +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
/* Apply the truncate. */
ReorderBufferApplyTruncate(rb, txn, nrelations,
relations, change,
streaming);
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
for (i = 0; i < nrelations; i++)
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
RelationClose(relations[i]);
Post-feature-freeze pgindent run. Discussion: https://postgr.es/m/15719.1523984266@sss.pgh.pa.us
2018-04-26 20:47:16 +02:00
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
break;
}
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
case REORDER_BUFFER_CHANGE_MESSAGE:
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
ReorderBufferApplyMessage(rb, txn, change, streaming);
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
break;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
case REORDER_BUFFER_CHANGE_INVALIDATION:
/* Execute the invalidation messages locally */
Clean up newlines following left parentheses Like commit c9d297751959.
2022-05-13 23:52:35 +02:00
ReorderBufferExecuteInvalidations(change->data.inval.ninvalidations,
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
change->data.inval.invalidations);
break;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT:
/* get rid of the old */
TeardownHistoricSnapshot(false);
if (snapshot_now->copied)
{
ReorderBufferFreeSnap(rb, snapshot_now);
snapshot_now =
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferCopySnap(rb, change->data.snapshot,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
txn, command_id);
}
pgindent run for 9.4 This includes removing tabs after periods in C comments, which was applied to back branches, so this change should not effect backpatching.
2014-05-06 18:12:18 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Restored from disk, need to be careful not to double
* free. We could introduce refcounting for that, but for
* now this seems infrequent enough not to care.
*/
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
else if (change->data.snapshot->copied)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
snapshot_now =
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferCopySnap(rb, change->data.snapshot,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
txn, command_id);
}
else
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
snapshot_now = change->data.snapshot;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/* and continue with the new one */
SetupHistoricSnapshot(snapshot_now, txn->tuplecid_hash);
break;
case REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID:
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(change->data.command_id != InvalidCommandId);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (command_id < change->data.command_id)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
command_id = change->data.command_id;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (!snapshot_now->copied)
{
/* we don't use the global one anymore */
snapshot_now = ReorderBufferCopySnap(rb, snapshot_now,
txn, command_id);
}
snapshot_now->curcid = command_id;
TeardownHistoricSnapshot(false);
SetupHistoricSnapshot(snapshot_now, txn->tuplecid_hash);
}
break;
case REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID:
elog(ERROR, "tuplecid value in changequeue");
break;
}
Fix the logical replication timeout during large DDLs. The DDLs like Refresh Materialized views that generate lots of temporary data due to rewrite rules may not be processed by output plugins (for example pgoutput). So, we won't send keep-alive messages for a long time while processing such commands and that can lead the subscriber side to timeout. We have previously fixed a similar case for large transactions in commit f95d53eded where the output plugin filters all or most of the changes but missed to handle the DDLs. We decided not to backpatch this as this adds a new callback in the existing exposed structure and moreover, users can increase the wal_sender_timeout and wal_receiver_timeout to avoid this problem. Author: Wang wei, Hou Zhijie Reviewed-by: Peter Smith, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/OS3PR01MB6275478E5D29E4A563302D3D9E2B9@OS3PR01MB6275.jpnprd01.prod.outlook.com Discussion: https://postgr.es/m/CAA5-nLARN7-3SLU_QUxfy510pmrYK6JJb=bk3hcgemAM_pAv+w@mail.gmail.com
2023-02-08 03:28:25 +01:00
/*
* It is possible that the data is not sent to downstream for a
* long time either because the output plugin filtered it or there
* is a DDL that generates a lot of data that is not processed by
* the plugin. So, in such cases, the downstream can timeout. To
* avoid that we try to send a keepalive message if required.
* Trying to send a keepalive message after every change has some
* overhead, but testing showed there is no noticeable overhead if
* we do it after every ~100 changes.
*/
#define CHANGES_THRESHOLD 100
if (++changes_count >= CHANGES_THRESHOLD)
{
rb->update_progress_txn(rb, txn, change->lsn);
changes_count = 0;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
/* speculative insertion record must be freed by now */
Assert(!specinsert);
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
Fix unsafe coding in ReorderBufferCommit(). "iterstate" must be marked volatile since it's changed inside the PG_TRY block and then used in the PG_CATCH stanza. Noted by Mark Wilding of Salesforce. (We really need to see if we can't get the C compiler to warn about this.) Also, reset iterstate to NULL after the mainline ReorderBufferIterTXNFinish call, to ensure the PG_CATCH block doesn't try to do that a second time.
2015-01-24 19:25:19 +01:00
/* clean up the iterator */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferIterTXNFinish(rb, iterstate);
Fix unsafe coding in ReorderBufferCommit(). "iterstate" must be marked volatile since it's changed inside the PG_TRY block and then used in the PG_CATCH stanza. Noted by Mark Wilding of Salesforce. (We really need to see if we can't get the C compiler to warn about this.) Also, reset iterstate to NULL after the mainline ReorderBufferIterTXNFinish call, to ensure the PG_CATCH block doesn't try to do that a second time.
2015-01-24 19:25:19 +01:00
iterstate = NULL;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +02:00
/*
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
* Update total transaction count and total bytes processed by the
* transaction and its subtransactions. Ensure to not count the
* streamed transaction multiple times.
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +02:00
*
* Note that the statistics computation has to be done after
* ReorderBufferIterTXNFinish as it releases the serialized change
* which we have already accounted in ReorderBufferIterTXNNext.
*/
if (!rbtxn_is_streamed(txn))
rb->totalTxns++;
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
rb->totalBytes += txn->total_size;
Add information of total data processed to replication slot stats. This adds the statistics about total transactions count and total transaction data logically sent to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats. Suggested-by: Andres Freund Author: Vignesh C and Amit Kapila Reviewed-by: Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-04-16 04:04:43 +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
/*
* Done with current changes, send the last message for this set of
* changes depending upon streaming mode.
*/
if (streaming)
{
if (stream_started)
{
rb->stream_stop(rb, txn, prev_lsn);
stream_started = false;
}
}
else
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
{
/*
* Call either PREPARE (for two-phase transactions) or COMMIT (for
* regular ones).
*/
if (rbtxn_prepared(txn))
rb->prepare(rb, txn, commit_lsn);
else
rb->commit(rb, txn, commit_lsn);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* this is just a sanity check against bad output plugin behaviour */
if (GetCurrentTransactionIdIfAny() != InvalidTransactionId)
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
elog(ERROR, "output plugin used XID %u",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
GetCurrentTransactionId());
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
/*
* Remember the command ID and snapshot for the next set of changes in
* streaming mode.
*/
if (streaming)
ReorderBufferSaveTXNSnapshot(rb, txn, snapshot_now, command_id);
else if (snapshot_now->copied)
ReorderBufferFreeSnap(rb, snapshot_now);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* cleanup */
TeardownHistoricSnapshot(false);
/*
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
* Aborting the current (sub-)transaction as a whole has the right
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* semantics. We want all locks acquired in here to be released, not
* reassigned to the parent and we do not want any database access
* have persistent effects.
*/
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
AbortCurrentTransaction();
/* make sure there's no cache pollution */
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
ReorderBufferExecuteInvalidations(txn->ninvalidations, txn->invalidations);
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
if (using_subtxn)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
RollbackAndReleaseCurrentSubTransaction();
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
/*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* We are here due to one of the four reasons: 1. Decoding an
* in-progress txn. 2. Decoding a prepared txn. 3. Decoding of a
* prepared txn that was (partially) streamed. 4. Decoding a committed
* txn.
*
* For 1, we allow truncation of txn data by removing the changes
* already streamed but still keeping other things like invalidations,
* snapshot, and tuplecids. For 2 and 3, we indicate
* ReorderBufferTruncateTXN to do more elaborate truncation of txn
* data as the entire transaction has been decoded except for commit.
* For 4, as the entire txn has been decoded, we can fully clean up
* the TXN reorder buffer.
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
*/
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
if (streaming || rbtxn_prepared(txn))
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
{
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
ReorderBufferTruncateTXN(rb, txn, rbtxn_prepared(txn));
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 the CheckXidAlive */
CheckXidAlive = InvalidTransactionId;
}
else
ReorderBufferCleanupTXN(rb, txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
PG_CATCH();
{
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
MemoryContext ecxt = MemoryContextSwitchTo(ccxt);
ErrorData *errdata = CopyErrorData();
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* TODO: Encapsulate cleanup from the PG_TRY and PG_CATCH blocks */
if (iterstate)
ReorderBufferIterTXNFinish(rb, iterstate);
TeardownHistoricSnapshot(true);
/*
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
* Force cache invalidation to happen outside of a valid transaction
* to prevent catalog access as we just caught an error.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
AbortCurrentTransaction();
/* make sure there's no cache pollution */
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
ReorderBufferExecuteInvalidations(txn->ninvalidations,
txn->invalidations);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
if (using_subtxn)
RollbackAndReleaseCurrentSubTransaction();
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
/*
* The error code ERRCODE_TRANSACTION_ROLLBACK indicates a concurrent
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* abort of the (sub)transaction we are streaming or preparing. We
* need to do the cleanup and return gracefully on this error, see
* SetupCheckXidLive.
Tighten the concurrent abort check during decoding. During decoding of an in-progress or prepared transaction, we detect concurrent abort with an error code ERRCODE_TRANSACTION_ROLLBACK. That is not sufficient because a callback can decide to throw that error code at other times as well. Reported-by: Tom Lane Author: Amit Kapila Reviewed-by: Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1KCjPRS4aZHB48QMM4J8XOC1+TD8jo-4Yu84E+MjwqVhA@mail.gmail.com
2021-05-06 04:56:42 +02:00
*
* This error code can be thrown by one of the callbacks we call
* during decoding so we need to ensure that we return gracefully only
* when we are sending the data in streaming mode and the streaming is
* not finished yet or when we are sending the data out on a PREPARE
* during a two-phase commit.
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
*/
Tighten the concurrent abort check during decoding. During decoding of an in-progress or prepared transaction, we detect concurrent abort with an error code ERRCODE_TRANSACTION_ROLLBACK. That is not sufficient because a callback can decide to throw that error code at other times as well. Reported-by: Tom Lane Author: Amit Kapila Reviewed-by: Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1KCjPRS4aZHB48QMM4J8XOC1+TD8jo-4Yu84E+MjwqVhA@mail.gmail.com
2021-05-06 04:56:42 +02:00
if (errdata->sqlerrcode == ERRCODE_TRANSACTION_ROLLBACK &&
(stream_started || rbtxn_prepared(txn)))
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
{
Tighten the concurrent abort check during decoding. During decoding of an in-progress or prepared transaction, we detect concurrent abort with an error code ERRCODE_TRANSACTION_ROLLBACK. That is not sufficient because a callback can decide to throw that error code at other times as well. Reported-by: Tom Lane Author: Amit Kapila Reviewed-by: Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1KCjPRS4aZHB48QMM4J8XOC1+TD8jo-4Yu84E+MjwqVhA@mail.gmail.com
2021-05-06 04:56:42 +02:00
/* curtxn must be set for streaming or prepared transactions */
Assert(curtxn);
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
/* Cleanup the temporary error state. */
FlushErrorState();
FreeErrorData(errdata);
errdata = NULL;
curtxn->concurrent_abort = true;
/* Reset the TXN so that it is allowed to stream remaining data. */
ReorderBufferResetTXN(rb, txn, snapshot_now,
command_id, prev_lsn,
specinsert);
}
else
{
ReorderBufferCleanupTXN(rb, txn);
MemoryContextSwitchTo(ecxt);
PG_RE_THROW();
}
}
PG_END_TRY();
}
Fix and improve cache invalidation logic for logical decoding. There are basically three situations in which logical decoding needs to perform cache invalidation. During/After replaying a transaction with catalog changes, when skipping a uninteresting transaction that performed catalog changes and when erroring out while replaying a transaction. Unfortunately these three cases were all done slightly differently - partially because 8de3e410fa, which greatly simplifies matters, got committed in the midst of the development of logical decoding. The actually problematic case was when logical decoding skipped transaction commits (and thus processed invalidations). When used via the SQL interface cache invalidation could access the catalog - bad, because we didn't set up enough state to allow that correctly. It'd not be hard to setup sufficient state, but the simpler solution is to always perform cache invalidation outside a valid transaction. Also make the different cache invalidation cases look as similar as possible, to ease code review. This fixes the assertion failure reported by Antonin Houska in 53EE02D9.7040702@gmail.com. The presented testcase has been expanded into a regression test. Backpatch to 9.4, where logical decoding was introduced.
2014-11-13 19:06:43 +01:00
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
/*
* Perform the replay of a transaction and its non-aborted subtransactions.
*
* Subtransactions previously have to be processed by
* ReorderBufferCommitChild(), even if previously assigned to the toplevel
* transaction with ReorderBufferAssignChild.
*
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
* This interface is called once a prepare or toplevel commit is read for both
* streamed as well as non-streamed transactions.
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
*/
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
static void
ReorderBufferReplay(ReorderBufferTXN *txn,
ReorderBuffer *rb, TransactionId xid,
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
XLogRecPtr commit_lsn, XLogRecPtr end_lsn,
TimestampTz commit_time,
RepOriginId origin_id, XLogRecPtr origin_lsn)
{
Snapshot snapshot_now;
CommandId command_id = FirstCommandId;
txn->final_lsn = commit_lsn;
txn->end_lsn = end_lsn;
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
txn->xact_time.commit_time = commit_time;
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
txn->origin_id = origin_id;
txn->origin_lsn = origin_lsn;
/*
* If the transaction was (partially) streamed, we need to commit it in a
* 'streamed' way. That is, we first stream the remaining part of the
* transaction, and then invoke stream_commit message.
*
* Called after everything (origin ID, LSN, ...) is stored in the
* transaction to avoid passing that information directly.
*/
if (rbtxn_is_streamed(txn))
{
ReorderBufferStreamCommit(rb, txn);
return;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
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
/*
* If this transaction has no snapshot, it didn't make any changes to the
* database, so there's nothing to decode. Note that
* ReorderBufferCommitChild will have transferred any snapshots from
* subtransactions if there were any.
*/
if (txn->base_snapshot == NULL)
{
Assert(txn->ninvalidations == 0);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* Removing this txn before a commit might result in the computation
* of an incorrect restart_lsn. See SnapBuildProcessRunningXacts.
*/
if (!rbtxn_prepared(txn))
ReorderBufferCleanupTXN(rb, txn);
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
return;
}
snapshot_now = txn->base_snapshot;
/* Process and send the changes to output plugin. */
ReorderBufferProcessTXN(rb, txn, commit_lsn, snapshot_now,
command_id, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* Commit a transaction.
*
* See comments for ReorderBufferReplay().
*/
void
ReorderBufferCommit(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr commit_lsn, XLogRecPtr end_lsn,
TimestampTz commit_time,
RepOriginId origin_id, XLogRecPtr origin_lsn)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
/* unknown transaction, nothing to replay */
if (txn == NULL)
return;
ReorderBufferReplay(txn, rb, xid, commit_lsn, end_lsn, commit_time,
origin_id, origin_lsn);
}
/*
* Record the prepare information for a transaction.
*/
bool
ReorderBufferRememberPrepareInfo(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr prepare_lsn, XLogRecPtr end_lsn,
TimestampTz prepare_time,
RepOriginId origin_id, XLogRecPtr origin_lsn)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr, false);
/* unknown transaction, nothing to do */
if (txn == NULL)
return false;
/*
* Remember the prepare information to be later used by commit prepared in
* case we skip doing prepare.
*/
txn->final_lsn = prepare_lsn;
txn->end_lsn = end_lsn;
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
txn->xact_time.prepare_time = prepare_time;
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
txn->origin_id = origin_id;
txn->origin_lsn = origin_lsn;
return true;
}
/* Remember that we have skipped prepare */
void
ReorderBufferSkipPrepare(ReorderBuffer *rb, TransactionId xid)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr, false);
/* unknown transaction, nothing to do */
if (txn == NULL)
return;
txn->txn_flags |= RBTXN_SKIPPED_PREPARE;
}
/*
* Prepare a two-phase transaction.
*
* See comments for ReorderBufferReplay().
*/
void
ReorderBufferPrepare(ReorderBuffer *rb, TransactionId xid,
char *gid)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
/* unknown transaction, nothing to replay */
if (txn == NULL)
return;
txn->txn_flags |= RBTXN_PREPARE;
txn->gid = pstrdup(gid);
/* The prepare info must have been updated in txn by now. */
Assert(txn->final_lsn != InvalidXLogRecPtr);
ReorderBufferReplay(txn, rb, xid, txn->final_lsn, txn->end_lsn,
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
txn->xact_time.prepare_time, txn->origin_id, txn->origin_lsn);
Ensure to send a prepare after we detect concurrent abort during decoding. It is possible that while decoding a prepared transaction, it gets aborted concurrently via a ROLLBACK PREPARED command. In that case, we were skipping all the changes and directly sending Rollback Prepared when we find the same in WAL. However, the downstream has no idea of the GID of such a transaction. So, ensure to send prepare even when a concurrent abort is detected. Author: Ajin Cherian Reviewed-by: Markus Wanner, Amit Kapila Discussion: https://postgr.es/m/f82133c6-6055-b400-7922-97dae9f2b50b@enterprisedb.com
2021-04-01 04:27:34 +02:00
/*
* We send the prepare for the concurrently aborted xacts so that later
* when rollback prepared is decoded and sent, the downstream should be
* able to rollback such a xact. See comments atop DecodePrepare.
Avoid sending prepare multiple times while decoding. We send the prepare for the concurrently aborted xacts so that later when rollback prepared is decoded and sent, the downstream should be able to rollback such a xact. For 'streaming' case (when we send changes for in-progress transactions), we were sending prepare twice when concurrent abort was detected. Author: Peter Smith Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/f82133c6-6055-b400-7922-97dae9f2b50b@enterprisedb.com
2021-04-26 07:57:44 +02:00
*
* Note, for the concurrent_abort + streaming case a stream_prepare was
* already sent within the ReorderBufferReplay call above.
Ensure to send a prepare after we detect concurrent abort during decoding. It is possible that while decoding a prepared transaction, it gets aborted concurrently via a ROLLBACK PREPARED command. In that case, we were skipping all the changes and directly sending Rollback Prepared when we find the same in WAL. However, the downstream has no idea of the GID of such a transaction. So, ensure to send prepare even when a concurrent abort is detected. Author: Ajin Cherian Reviewed-by: Markus Wanner, Amit Kapila Discussion: https://postgr.es/m/f82133c6-6055-b400-7922-97dae9f2b50b@enterprisedb.com
2021-04-01 04:27:34 +02:00
*/
Avoid sending prepare multiple times while decoding. We send the prepare for the concurrently aborted xacts so that later when rollback prepared is decoded and sent, the downstream should be able to rollback such a xact. For 'streaming' case (when we send changes for in-progress transactions), we were sending prepare twice when concurrent abort was detected. Author: Peter Smith Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/f82133c6-6055-b400-7922-97dae9f2b50b@enterprisedb.com
2021-04-26 07:57:44 +02:00
if (txn->concurrent_abort && !rbtxn_is_streamed(txn))
Ensure to send a prepare after we detect concurrent abort during decoding. It is possible that while decoding a prepared transaction, it gets aborted concurrently via a ROLLBACK PREPARED command. In that case, we were skipping all the changes and directly sending Rollback Prepared when we find the same in WAL. However, the downstream has no idea of the GID of such a transaction. So, ensure to send prepare even when a concurrent abort is detected. Author: Ajin Cherian Reviewed-by: Markus Wanner, Amit Kapila Discussion: https://postgr.es/m/f82133c6-6055-b400-7922-97dae9f2b50b@enterprisedb.com
2021-04-01 04:27:34 +02:00
rb->prepare(rb, txn, txn->final_lsn);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
}
/*
* This is used to handle COMMIT/ROLLBACK PREPARED.
*/
void
ReorderBufferFinishPrepared(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr commit_lsn, XLogRecPtr end_lsn,
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
XLogRecPtr two_phase_at,
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
TimestampTz commit_time, RepOriginId origin_id,
XLogRecPtr origin_lsn, char *gid, bool is_commit)
{
ReorderBufferTXN *txn;
XLogRecPtr prepare_end_lsn;
TimestampTz prepare_time;
Fix an oversight in ReorderBufferFinishPrepared. We don't have anything to decode in a transaction if ReorderBufferTXN doesn't exist by the time we decode the commit prepared. So don't create a new ReorderBufferTXN here. This is an oversight in commit a271a1b5. Reported-by: Markus Wanner Discussion: https://postgr.es/m/dbec82e2-dbd7-95a2-c6b6-e488cbbdf853@bluegap.ch
2021-02-23 05:17:41 +01:00
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, commit_lsn, false);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/* unknown transaction, nothing to do */
if (txn == NULL)
return;
/*
* By this time the txn has the prepare record information, remember it to
* be later used for rollback.
*/
prepare_end_lsn = txn->end_lsn;
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
prepare_time = txn->xact_time.prepare_time;
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/* add the gid in the txn */
txn->gid = pstrdup(gid);
/*
* It is possible that this transaction is not decoded at prepare time
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
* either because by that time we didn't have a consistent snapshot, or
* two_phase was not enabled, or it was decoded earlier but we have
* restarted. We only need to send the prepare if it was not decoded
* earlier. We don't need to decode the xact for aborts if it is not done
* already.
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
*/
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
if ((txn->final_lsn < two_phase_at) && is_commit)
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
{
txn->txn_flags |= RBTXN_PREPARE;
/*
* The prepare info must have been updated in txn even if we skip
* prepare.
*/
Assert(txn->final_lsn != InvalidXLogRecPtr);
/*
* By this time the txn has the prepare record information and it is
* important to use that so that downstream gets the accurate
* information. If instead, we have passed commit information here
* then downstream can behave as it has already replayed commit
* prepared after the restart.
*/
ReorderBufferReplay(txn, rb, xid, txn->final_lsn, txn->end_lsn,
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
txn->xact_time.prepare_time, txn->origin_id, txn->origin_lsn);
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
}
txn->final_lsn = commit_lsn;
txn->end_lsn = end_lsn;
Add support for prepared transactions to built-in logical replication. To add support for streaming transactions at prepare time into the built-in logical replication, we need to do the following things: * Modify the output plugin (pgoutput) to implement the new two-phase API callbacks, by leveraging the extended replication protocol. * Modify the replication apply worker, to properly handle two-phase transactions by replaying them on prepare. * Add a new SUBSCRIPTION option "two_phase" to allow users to enable two-phase transactions. We enable the two_phase once the initial data sync is over. We however must explicitly disable replication of two-phase transactions during replication slot creation, even if the plugin supports it. We don't need to replicate the changes accumulated during this phase, and moreover, we don't have a replication connection open so we don't know where to send the data anyway. The streaming option is not allowed with this new two_phase option. This can be done as a separate patch. We don't allow to toggle two_phase option of a subscription because it can lead to an inconsistent replica. For the same reason, we don't allow to refresh the publication once the two_phase is enabled for a subscription unless copy_data option is false. Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow Tested-By: Haiying Tang Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
2021-07-14 04:03:50 +02:00
txn->xact_time.commit_time = commit_time;
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
txn->origin_id = origin_id;
txn->origin_lsn = origin_lsn;
if (is_commit)
rb->commit_prepared(rb, txn, commit_lsn);
else
rb->rollback_prepared(rb, txn, prepare_end_lsn, prepare_time);
/* cleanup: make sure there's no cache pollution */
ReorderBufferExecuteInvalidations(txn->ninvalidations,
txn->invalidations);
ReorderBufferCleanupTXN(rb, txn);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Abort a transaction that possibly has previous changes. Needs to be first
* called for subtransactions and then for the toplevel xid.
*
* NB: Transactions handled here have to have actively aborted (i.e. have
* produced an abort record). Implicitly aborted transactions are handled via
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
* ReorderBufferAbortOld(); transactions we're just not interested in, but
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* which have committed are handled in ReorderBufferForget().
*
* This function purges this transaction and its contents from memory and
* disk.
*/
void
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
ReorderBufferAbort(ReorderBuffer *rb, TransactionId xid, XLogRecPtr lsn,
TimestampTz abort_time)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
/* unknown, nothing to remove */
if (txn == NULL)
return;
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
txn->xact_time.abort_time = abort_time;
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
/* For streamed transactions notify the remote node about the abort. */
if (rbtxn_is_streamed(txn))
{
rb->stream_abort(rb, txn, lsn);
/*
* We might have decoded changes for this transaction that could load
* the cache as per the current transaction's view (consider DDL's
* happened in this transaction). We don't want the decoding of future
* transactions to use those cache entries so execute invalidations.
*/
if (txn->ninvalidations > 0)
ReorderBufferImmediateInvalidation(rb, txn->ninvalidations,
txn->invalidations);
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* cosmetic... */
txn->final_lsn = lsn;
/* remove potential on-disk data, and deallocate */
ReorderBufferCleanupTXN(rb, txn);
}
/*
* Abort all transactions that aren't actually running anymore because the
* server restarted.
*
* NB: These really have to be transactions that have aborted due to a server
* crash/immediate restart, as we don't deal with invalidations here.
*/
void
ReorderBufferAbortOld(ReorderBuffer *rb, TransactionId oldestRunningXid)
{
dlist_mutable_iter it;
/*
* Iterate through all (potential) toplevel TXNs and abort all that are
* older than what possibly can be running. Once we've found the first
* that is alive we stop, there might be some that acquired an xid earlier
Fix failure to delete spill files of aborted transactions Logical decoding's reorderbuffer.c may spill transaction files to disk when transactions are large. These are supposed to be removed when they become "too old" by xid; but file removal requires the boundary LSNs of the transaction to be known. The final_lsn is only set when we see the commit or abort record for the transaction, but nothing sets the value for transactions that crash, so the removal code misbehaves -- in assertion-enabled builds, it crashes by a failed assertion. To fix, modify the final_lsn of transactions that don't have a value set, to the LSN of the very latest change in the transaction. This causes the spilled files to be removed appropriately. Author: Atsushi Torikoshi Reviewed-by: Kyotaro HORIGUCHI, Craig Ringer, Masahiko Sawada Discussion: https://postgr.es/m/54e4e488-186b-a056-6628-50628e4e4ebc@lab.ntt.co.jp
2018-01-05 16:17:10 +01:00
* but started writing later, but it's unlikely and they will be cleaned
* up in a later call to this function.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
dlist_foreach_modify(it, &rb->toplevel_by_lsn)
{
ReorderBufferTXN *txn;
txn = dlist_container(ReorderBufferTXN, node, it.cur);
if (TransactionIdPrecedes(txn->xid, oldestRunningXid))
{
Fix race condition leading to hanging logical slot creation. The snapshot assembly during the creation of logical slots relied waiting for transactions in xl_running_xacts to end, by checking for their commit/abort records. Unfortunately, despite locking, it is possible to see an xl_running_xact record listing transactions as ready, that have already WAL-logged an commit/abort record, as the locking just prevents the ProcArray to be adjusted, and the commit record has to be logged first. That lead to either delayed or hanging snapshot creation, because snapbuild.c would wait "forever" to see commit/abort records for some transactions. That hang resolved only if a xl_running_xacts record without any running transactions happened to be logged, far from certain on a busy server. It's impractical to prevent that via more heavyweight locking, the likelihood of deadlocks and significantly increased contention would be too big. Instead change the initial snapshot creation to be solely based on tracking the oldest running transaction via xl_running_xacts->oldestRunningXid - that actually ends up significantly simplifying the code. That has two disadvantages: 1) Because we cannot fully "trust" the contents of xl_running_xacts, we cannot use it to build the initial snapshot. Instead we have to wait twice for all running transactions to finish. 2) Previously a slot, unless the race occurred, could be created when the all transaction perceived as running based on commit/abort records, now we have to wait for the next xl_running_xacts record. To address that, trigger logging new xl_running_xacts record from within snapbuild.c exactly when necessary. Unfortunately snabuild.c's SnapBuild is stored on disk, one of the stupider ideas of a certain Mr Freund, so we can't change it in a minor release. As this is going to be backpatched, we have to hack around a bit to keep on-disk compatibility. A later commit will rejigger that on master. Author: Andres Freund, based on a quite different patch from Petr Jelinek Analyzed-By: Petr Jelinek Reviewed-By: Petr Jelinek Discussion: https://postgr.es/m/f37e975c-908f-858e-707f-058d3b1eb214@2ndquadrant.com Backpatch: 9.4-, where logical decoding has been introduced
2017-05-13 23:21:00 +02:00
elog(DEBUG2, "aborting old transaction %u", txn->xid);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Remove the streaming files for incomplete xacts after restart. After restart, we try to stream the changes for large transactions that were not sent before server crash and restart. However, we forget to send the abort message for such transactions. This leads to spurious streaming files on the subscriber which won't be cleaned till the apply worker or the subscriber server restarts. Reported-by: Dilip Kumar Author: Hou Zhijie Reviewed-by: Dilip Kumar and Amit Kapila Backpatch-through: 14 Discussion: https://postgr.es/m/OS0PR01MB5716A773F46768A1B75BE24394FB9@OS0PR01MB5716.jpnprd01.prod.outlook.com
2023-01-07 07:47:14 +01:00
/* Notify the remote node about the crash/immediate restart. */
if (rbtxn_is_streamed(txn))
rb->stream_abort(rb, txn, InvalidXLogRecPtr);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* remove potential on-disk data, and deallocate this tx */
ReorderBufferCleanupTXN(rb, txn);
}
else
return;
}
}
/*
Fix typos. Reviewed-by: Michael Paquier Discussion: https://postgr.es/m/CA%2BhUKGJFWXmtYo6Frd77RR8YXCHz7hJ2mRy5aHV%3D7fJOqDnBHA%40mail.gmail.com
2019-05-23 03:17:41 +02:00
* Forget the contents of a transaction if we aren't interested in its
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* contents. Needs to be first called for subtransactions and then for the
* toplevel xid.
*
* This is significantly different to ReorderBufferAbort() because
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
* transactions that have committed need to be treated differently from aborted
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* ones since they may have modified the catalog.
*
* Note that this is only allowed to be called in the moment a transaction
* commit has just been read, not earlier; otherwise later records referring
* to this xid might re-create the transaction incompletely.
*/
void
ReorderBufferForget(ReorderBuffer *rb, TransactionId xid, XLogRecPtr lsn)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
/* unknown, nothing to forget */
if (txn == NULL)
return;
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
/* this transaction mustn't be streamed */
Assert(!rbtxn_is_streamed(txn));
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
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* cosmetic... */
txn->final_lsn = lsn;
/*
Fix more typos in comments. Patch by CharSyam, plus a few more I spotted with grep.
2015-05-20 18:44:46 +02:00
* Process cache invalidation messages if there are any. Even if we're not
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* interested in the transaction's contents, it could have manipulated the
* catalog and we need to update the caches according to that.
*/
if (txn->base_snapshot != NULL && txn->ninvalidations > 0)
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
ReorderBufferImmediateInvalidation(rb, txn->ninvalidations,
txn->invalidations);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
else
Assert(txn->ninvalidations == 0);
/* remove potential on-disk data, and deallocate */
ReorderBufferCleanupTXN(rb, txn);
}
Allow decoding at prepare time in ReorderBuffer. This patch allows PREPARE-time decoding of two-phase transactions (if the output plugin supports this capability), in which case the transactions are replayed at PREPARE and then committed later when COMMIT PREPARED arrives. Now that we decode the changes before the commit, the concurrent aborts may cause failures when the output plugin consults catalogs (both system and user-defined). We detect such failures with a special sqlerrcode ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a6e and stop decoding the remaining changes. Then we rollback the changes when rollback prepared is encountered. Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar Tested-by: Takamichi Osumi Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
2021-01-04 04:04:50 +01:00
/*
* Invalidate cache for those transactions that need to be skipped just in case
* catalogs were manipulated as part of the transaction.
*
* Note that this is a special-purpose function for prepared transactions where
* we don't want to clean up the TXN even when we decide to skip it. See
* DecodePrepare.
*/
void
ReorderBufferInvalidate(ReorderBuffer *rb, TransactionId xid, XLogRecPtr lsn)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
/* unknown, nothing to do */
if (txn == NULL)
return;
/*
* Process cache invalidation messages if there are any. Even if we're not
* interested in the transaction's contents, it could have manipulated the
* catalog and we need to update the caches according to that.
*/
if (txn->base_snapshot != NULL && txn->ninvalidations > 0)
ReorderBufferImmediateInvalidation(rb, txn->ninvalidations,
txn->invalidations);
else
Assert(txn->ninvalidations == 0);
}
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
/*
* Execute invalidations happening outside the context of a decoded
* transaction. That currently happens either for xid-less commits
Fix Latin spelling "c.f." should be "cf.".
2018-01-11 14:31:11 +01:00
* (cf. RecordTransactionCommit()) or for invalidations in uninteresting
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 (via ReorderBufferForget()).
*/
void
ReorderBufferImmediateInvalidation(ReorderBuffer *rb, uint32 ninvalidations,
SharedInvalidationMessage *invalidations)
{
bool use_subtxn = IsTransactionOrTransactionBlock();
int i;
if (use_subtxn)
BeginInternalSubTransaction("replay");
/*
* Force invalidations to happen outside of a valid transaction - that way
* entries will just be marked as invalid without accessing the catalog.
* That's advantageous because we don't need to setup the full state
* necessary for catalog access.
*/
if (use_subtxn)
AbortCurrentTransaction();
for (i = 0; i < ninvalidations; i++)
LocalExecuteInvalidationMessage(&invalidations[i]);
if (use_subtxn)
RollbackAndReleaseCurrentSubTransaction();
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
logical decoding: Tell reorderbuffer about all xids. Logical decoding's reorderbuffer keeps transactions in an LSN ordered list for efficiency. To make that's efficiently possible upper-level xids are forced to be logged before nested subtransaction xids. That only works though if these records are all looked at: Unfortunately we didn't do so for e.g. row level locks, which are otherwise uninteresting for logical decoding. This could lead to errors like: "ERROR: subxact logged without previous toplevel record". It's not sufficient to just look at row locking records, the xid could appear first due to a lot of other types of records (which will trigger the transaction to be marked logged with MarkCurrentTransactionIdLoggedIfAny). So invent infrastructure to tell reorderbuffer about xids seen, when they'd otherwise not pass through reorderbuffer.c. Reported-By: Jarred Ward Bug: #13844 Discussion: 20160105033249.1087.66040@wrigleys.postgresql.org Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
* Tell reorderbuffer about an xid seen in the WAL stream. Has to be called at
* least once for every xid in XLogRecord->xl_xid (other places in records
* may, but do not have to be passed through here).
*
* Reorderbuffer keeps some datastructures about transactions in LSN order,
* for efficiency. To do that it has to know about when transactions are seen
* first in the WAL. As many types of records are not actually interesting for
* logical decoding, they do not necessarily pass though here.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
logical decoding: Tell reorderbuffer about all xids. Logical decoding's reorderbuffer keeps transactions in an LSN ordered list for efficiency. To make that's efficiently possible upper-level xids are forced to be logged before nested subtransaction xids. That only works though if these records are all looked at: Unfortunately we didn't do so for e.g. row level locks, which are otherwise uninteresting for logical decoding. This could lead to errors like: "ERROR: subxact logged without previous toplevel record". It's not sufficient to just look at row locking records, the xid could appear first due to a lot of other types of records (which will trigger the transaction to be marked logged with MarkCurrentTransactionIdLoggedIfAny). So invent infrastructure to tell reorderbuffer about xids seen, when they'd otherwise not pass through reorderbuffer.c. Reported-By: Jarred Ward Bug: #13844 Discussion: 20160105033249.1087.66040@wrigleys.postgresql.org Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
void
ReorderBufferProcessXid(ReorderBuffer *rb, TransactionId xid, XLogRecPtr lsn)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
logical decoding: Tell reorderbuffer about all xids. Logical decoding's reorderbuffer keeps transactions in an LSN ordered list for efficiency. To make that's efficiently possible upper-level xids are forced to be logged before nested subtransaction xids. That only works though if these records are all looked at: Unfortunately we didn't do so for e.g. row level locks, which are otherwise uninteresting for logical decoding. This could lead to errors like: "ERROR: subxact logged without previous toplevel record". It's not sufficient to just look at row locking records, the xid could appear first due to a lot of other types of records (which will trigger the transaction to be marked logged with MarkCurrentTransactionIdLoggedIfAny). So invent infrastructure to tell reorderbuffer about xids seen, when they'd otherwise not pass through reorderbuffer.c. Reported-By: Jarred Ward Bug: #13844 Discussion: 20160105033249.1087.66040@wrigleys.postgresql.org Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
/* many records won't have an xid assigned, centralize check here */
if (xid != InvalidTransactionId)
ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Add a new snapshot to this transaction that may only used after lsn 'lsn'
* because the previous snapshot doesn't describe the catalog correctly for
* following rows.
*/
void
ReorderBufferAddSnapshot(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr lsn, Snapshot snap)
{
ReorderBufferChange *change = ReorderBufferGetChange(rb);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
change->data.snapshot = snap;
change->action = REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
ReorderBufferQueueChange(rb, xid, lsn, change, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
* Set up the transaction's base snapshot.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
* If we know that xid is a subtransaction, set the base snapshot on the
* top-level transaction instead.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
ReorderBufferSetBaseSnapshot(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr lsn, Snapshot snap)
{
ReorderBufferTXN *txn;
bool is_new;
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(snap != NULL);
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/*
* Fetch the transaction to operate on. If we know it's a subtransaction,
* operate on its top-level transaction instead.
*/
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
txn = ReorderBufferTXNByXid(rb, xid, true, &is_new, lsn, true);
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_known_subxact(txn))
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
txn = ReorderBufferTXNByXid(rb, txn->toplevel_xid, false,
NULL, InvalidXLogRecPtr, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(txn->base_snapshot == NULL);
txn->base_snapshot = snap;
txn->base_snapshot_lsn = lsn;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
dlist_push_tail(&rb->txns_by_base_snapshot_lsn, &txn->base_snapshot_node);
AssertTXNLsnOrder(rb);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Access the catalog with this CommandId at this point in the changestream.
*
* May only be called for command ids > 1
*/
void
ReorderBufferAddNewCommandId(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr lsn, CommandId cid)
{
ReorderBufferChange *change = ReorderBufferGetChange(rb);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
change->data.command_id = cid;
change->action = REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
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
ReorderBufferQueueChange(rb, xid, lsn, change, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
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
* Update memory counters to account for the new or removed change.
*
* We update two counters - in the reorder buffer, and in the transaction
* containing the change. The reorder buffer counter allows us to quickly
* decide if we reached the memory limit, the transaction counter allows
* us to quickly pick the largest transaction for eviction.
*
* When streaming is enabled, we need to update the toplevel transaction
* counters instead - we don't really care about subtransactions as we
* can't stream them individually anyway, and we only pick toplevel
* transactions for eviction. So only toplevel transactions matter.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*/
static void
ReorderBufferChangeMemoryUpdate(ReorderBuffer *rb,
ReorderBufferChange *change,
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
bool addition, Size sz)
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
{
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
ReorderBufferTXN *txn;
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
ReorderBufferTXN *toptxn;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Assert(change->txn);
/*
* Ignore tuple CID changes, because those are not evicted when reaching
* memory limit. So we just don't count them, because it might easily
* trigger a pointless attempt to spill.
*/
if (change->action == REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID)
return;
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
txn = change->txn;
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
/*
* Update the total size in top level as well. This is later used to
* compute the decoding stats.
*/
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
toptxn = rbtxn_get_toptxn(txn);
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
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
if (addition)
{
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
txn->size += sz;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
rb->size += sz;
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
/* Update the total size in the top transaction. */
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
toptxn->total_size += sz;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
}
else
{
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
Assert((rb->size >= sz) && (txn->size >= sz));
txn->size -= sz;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
rb->size -= sz;
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
/* Update the total size in the top transaction. */
Fix the computation of slot stats for 'total_bytes'. Previously, we were using the size of all the changes present in ReorderBuffer to compute total_bytes after decoding a transaction and that can lead to counting some of the transactions' changes more than once. Fix it by using the size of the changes decoded for a transaction to compute 'total_bytes'. Author: Sawada Masahiko Reviewed-by: Vignesh C, Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-03 03:52:08 +02:00
toptxn->total_size -= sz;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
}
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
Assert(txn->size <= rb->size);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +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
* Add new (relfilelocator, tid) -> (cmin, cmax) mappings.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*
* We do not include this change type in memory accounting, because we
* keep CIDs in a separate list and do not evict them when reaching
* the memory limit.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
ReorderBufferAddNewTupleCids(ReorderBuffer *rb, TransactionId xid,
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
XLogRecPtr lsn, RelFileLocator locator,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ItemPointerData tid, CommandId cmin,
CommandId cmax, CommandId combocid)
{
ReorderBufferChange *change = ReorderBufferGetChange(rb);
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
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
change->data.tuplecid.locator = locator;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
change->data.tuplecid.tid = tid;
change->data.tuplecid.cmin = cmin;
change->data.tuplecid.cmax = cmax;
change->data.tuplecid.combocid = combocid;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
change->lsn = lsn;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
change->txn = txn;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
change->action = REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
dlist_push_tail(&txn->tuplecids, &change->node);
txn->ntuplecids++;
}
/*
Fix comments atop ReorderBufferAddInvalidations. The comments atop seem to indicate that we always accumulate invalidation messages in a top-level transaction which is neither required nor matches with the code. Author: Amit Kapila Reviewd by: Masahiko Sawada Backpatch-through: 14, where it was introduced in commit c55040ccd0 Discussion: https://postgr.es/m/CAA4eK1LxGgnUroPz8STb6OfjVU1yaHoSA+T63URwmGCLdMJ0LA@mail.gmail.com
2022-11-10 12:26:49 +01:00
* Accumulate the invalidations for executing them later.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*
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
* This needs to be called for each XLOG_XACT_INVALIDATIONS message and
Fix comments atop ReorderBufferAddInvalidations. The comments atop seem to indicate that we always accumulate invalidation messages in a top-level transaction which is neither required nor matches with the code. Author: Amit Kapila Reviewd by: Masahiko Sawada Backpatch-through: 14, where it was introduced in commit c55040ccd0 Discussion: https://postgr.es/m/CAA4eK1LxGgnUroPz8STb6OfjVU1yaHoSA+T63URwmGCLdMJ0LA@mail.gmail.com
2022-11-10 12:26:49 +01:00
* accumulates all the invalidation messages in the toplevel transaction, if
* available, otherwise in the current transaction, as well as in the form of
* change in reorder buffer. We require to record it in form of the change
* so that we can execute only the required invalidations instead of executing
* all the invalidations on each CommandId increment. We also need to
* accumulate these in the txn buffer because in some cases where we skip
* processing the transaction (see ReorderBufferForget), we need to execute
* all the invalidations together.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
void
ReorderBufferAddInvalidations(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr lsn, Size nmsgs,
SharedInvalidationMessage *msgs)
{
ReorderBufferTXN *txn;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
MemoryContext oldcontext;
ReorderBufferChange *change;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
txn = ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
oldcontext = MemoryContextSwitchTo(rb->context);
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
/*
Fix comments atop ReorderBufferAddInvalidations. The comments atop seem to indicate that we always accumulate invalidation messages in a top-level transaction which is neither required nor matches with the code. Author: Amit Kapila Reviewd by: Masahiko Sawada Backpatch-through: 14, where it was introduced in commit c55040ccd0 Discussion: https://postgr.es/m/CAA4eK1LxGgnUroPz8STb6OfjVU1yaHoSA+T63URwmGCLdMJ0LA@mail.gmail.com
2022-11-10 12:26:49 +01:00
* Collect all the invalidations under the top transaction, if available,
* so that we can execute them all together. See comments atop this
* function.
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
*/
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
txn = rbtxn_get_toptxn(txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(nmsgs > 0);
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
/* Accumulate invalidations. */
if (txn->ninvalidations == 0)
{
txn->ninvalidations = nmsgs;
txn->invalidations = (SharedInvalidationMessage *)
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
palloc(sizeof(SharedInvalidationMessage) * nmsgs);
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
memcpy(txn->invalidations, msgs,
sizeof(SharedInvalidationMessage) * nmsgs);
}
else
{
txn->invalidations = (SharedInvalidationMessage *)
repalloc(txn->invalidations, sizeof(SharedInvalidationMessage) *
(txn->ninvalidations + nmsgs));
memcpy(txn->invalidations + txn->ninvalidations, msgs,
nmsgs * sizeof(SharedInvalidationMessage));
txn->ninvalidations += nmsgs;
}
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
change = ReorderBufferGetChange(rb);
change->action = REORDER_BUFFER_CHANGE_INVALIDATION;
change->data.inval.ninvalidations = nmsgs;
change->data.inval.invalidations = (SharedInvalidationMessage *)
palloc(sizeof(SharedInvalidationMessage) * nmsgs);
memcpy(change->data.inval.invalidations, msgs,
sizeof(SharedInvalidationMessage) * nmsgs);
ReorderBufferQueueChange(rb, xid, lsn, change, false);
MemoryContextSwitchTo(oldcontext);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Apply all invalidations we know. Possibly we only need parts at this point
* in the changestream but we don't know which those are.
*/
static void
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
ReorderBufferExecuteInvalidations(uint32 nmsgs, SharedInvalidationMessage *msgs)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
int i;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
for (i = 0; i < nmsgs; i++)
LocalExecuteInvalidationMessage(&msgs[i]);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Mark a transaction as containing catalog changes
*/
void
ReorderBufferXidSetCatalogChanges(ReorderBuffer *rb, TransactionId xid,
XLogRecPtr lsn)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, true, NULL, lsn, true);
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
if (!rbtxn_has_catalog_changes(txn))
{
txn->txn_flags |= RBTXN_HAS_CATALOG_CHANGES;
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
dclist_push_tail(&rb->catchange_txns, &txn->catchange_node);
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +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
/*
* Mark top-level transaction as having catalog changes too if one of its
* children has so that the ReorderBufferBuildTupleCidHash can
* conveniently check just top-level transaction and decide whether to
* build the hash table or not.
*/
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
if (rbtxn_is_subtxn(txn))
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
{
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
ReorderBufferTXN *toptxn = rbtxn_get_toptxn(txn);
if (!rbtxn_has_catalog_changes(toptxn))
{
toptxn->txn_flags |= RBTXN_HAS_CATALOG_CHANGES;
dclist_push_tail(&rb->catchange_txns, &toptxn->catchange_node);
}
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
}
}
/*
* Return palloc'ed array of the transactions that have changed catalogs.
* The returned array is sorted in xidComparator order.
*
* The caller must free the returned array when done with it.
*/
TransactionId *
ReorderBufferGetCatalogChangesXacts(ReorderBuffer *rb)
{
dlist_iter iter;
TransactionId *xids = NULL;
size_t xcnt = 0;
/* Quick return if the list is empty */
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
if (dclist_count(&rb->catchange_txns) == 0)
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
return NULL;
/* Initialize XID array */
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
xids = (TransactionId *) palloc(sizeof(TransactionId) *
dclist_count(&rb->catchange_txns));
dclist_foreach(iter, &rb->catchange_txns)
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
{
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
ReorderBufferTXN *txn = dclist_container(ReorderBufferTXN,
catchange_node,
iter.cur);
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
Assert(rbtxn_has_catalog_changes(txn));
xids[xcnt++] = txn->xid;
}
qsort(xids, xcnt, sizeof(TransactionId), xidComparator);
Add doubly linked count list implementation We have various requirements when using a dlist_head to keep track of the number of items in the list. This, traditionally, has been done by maintaining a counter variable in the calling code. Here we tidy this up by adding "dclist", which is very similar to dlist but also keeps track of the number of items stored in the list. Callers may use the new dclist_count() function when they need to know how many items are stored. Obtaining the count is an O(1) operation. For simplicity reasons, dclist and dlist both use dlist_node as their node type and dlist_iter/dlist_mutable_iter as their iterator type. dclists have all of the same functionality as dlists except there is no function named dclist_delete(). To remove an item from a list dclist_delete_from() must be used. This requires knowing which dclist the given item is stored in. Additionally, here we also convert some dlists where additional code exists to keep track of the number of items stored and to make these use dclists instead. Author: David Rowley Reviewed-by: Bharath Rupireddy, Aleksander Alekseev Discussion: https://postgr.es/m/CAApHDvrtVxr+FXEX0VbViCFKDGxA3tWDgw9oFewNXCJMmwLjLg@mail.gmail.com
2022-11-02 02:06:05 +01:00
Assert(xcnt == dclist_count(&rb->catchange_txns));
Fix catalog lookup with the wrong snapshot during logical decoding. Previously, we relied on HEAP2_NEW_CID records and XACT_INVALIDATION records to know if the transaction has modified the catalog, and that information is not serialized to snapshot. Therefore, after the restart, if the logical decoding decodes only the commit record of the transaction that has actually modified a catalog, we will miss adding its XID to the snapshot. Thus, we will end up looking at catalogs with the wrong snapshot. To fix this problem, this change adds the list of transaction IDs and sub-transaction IDs, that have modified catalogs and are running during snapshot serialization, to the serialized snapshot. After restart or otherwise, when we restore from such a serialized snapshot, the corresponding list is restored in memory. Now, when decoding a COMMIT record, we check both the list and the ReorderBuffer to see if the transaction has modified catalogs. Since this adds additional information to the serialized snapshot, we cannot backpatch it. For back branches, we took another approach. We remember the last-running-xacts list of the decoded RUNNING_XACTS record after restoring the previously serialized snapshot. Then, we mark the transaction as containing catalog changes if it's in the list of initial running transactions and its commit record has XACT_XINFO_HAS_INVALS. This doesn't require any file format changes but the transaction will end up being added to the snapshot even if it has only relcache invalidations. But that won't be a problem since we use snapshot built during decoding only to read system catalogs. This commit bumps SNAPBUILD_VERSION because of a change in SnapBuild. Reported-by: Mike Oh Author: Masahiko Sawada Reviewed-by: Amit Kapila, Shi yu, Takamichi Osumi, Kyotaro Horiguchi, Bertrand Drouvot, Ahsan Hadi Backpatch-through: 10 Discussion: https://postgr.es/m/81D0D8B0-E7C4-4999-B616-1E5004DBDCD2%40amazon.com
2022-08-11 06:39:24 +02:00
return xids;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Query whether a transaction is already *known* to contain catalog
* changes. This can be wrong until directly before the commit!
*/
bool
ReorderBufferXidHasCatalogChanges(ReorderBuffer *rb, TransactionId xid)
{
ReorderBufferTXN *txn;
txn = ReorderBufferTXNByXid(rb, xid, false, NULL, InvalidXLogRecPtr,
false);
if (txn == NULL)
return false;
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
return rbtxn_has_catalog_changes(txn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
* ReorderBufferXidHasBaseSnapshot
* Have we already set the base snapshot for the given txn/subtxn?
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
bool
ReorderBufferXidHasBaseSnapshot(ReorderBuffer *rb, TransactionId xid)
{
ReorderBufferTXN *txn;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
txn = ReorderBufferTXNByXid(rb, xid, false,
NULL, InvalidXLogRecPtr, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* transaction isn't known yet, ergo no snapshot */
if (txn == NULL)
return false;
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
/* a known subtxn? operate on top-level txn instead */
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
if (rbtxn_is_known_subxact(txn))
Fix "base" snapshot handling in logical decoding Two closely related bugs are fixed. First, xmin of logical slots was advanced too early. During xl_running_xacts processing, xmin of the slot was set to the oldest running xid in the record, but that's wrong: actually, snapshots which will be used for not-yet-replayed transactions might consider older txns as running too, so we need to keep xmin back for them. The problem wasn't noticed earlier because DDL which allows to delete tuple (set xmax) while some another not-yet-committed transaction looks at it is pretty rare, if not unique: e.g. all forms of ALTER TABLE which change schema acquire ACCESS EXCLUSIVE lock conflicting with any inserts. The included test case (test_decoding's oldest_xmin) uses ALTER of a composite type, which doesn't have such interlocking. To deal with this, we must be able to quickly retrieve oldest xmin (oldest running xid among all assigned snapshots) from ReorderBuffer. To fix, add another list of ReorderBufferTXNs to the reorderbuffer, where transactions are sorted by base-snapshot-LSN. This is slightly different from the existing (sorted by first-LSN) list, because a transaction can have an earlier LSN but a later Xmin, if its first record does not obtain an xmin (eg. xl_xact_assignment). Note this new list doesn't fully replace the existing txn list: we still need that one to prevent WAL recycling. The second issue concerns SnapBuilder snapshots and subtransactions. SnapBuildDistributeNewCatalogSnapshot never assigned a snapshot to a transaction that is known to be a subtxn, which is good in the common case that the top-level transaction already has one (no point in doing so), but a bug otherwise. To fix, arrange to transfer the snapshot from the subtxn to its top-level txn as soon as the kinship gets known. test_decoding's snapshot_transfer verifies this. Also, fix a minor memory leak: refcount of toplevel's old base snapshot was not decremented when the snapshot is transferred from child. Liberally sprinkle code comments, and rewrite a few existing ones. This part is my (Álvaro's) contribution to this commit, as I had to write all those comments in order to understand the existing code and Arseny's patch. Reported-by: Arseny Sher <a.sher@postgrespro.ru> Diagnosed-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Arseny Sher <a.sher@postgrespro.ru> Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Antonin Houska <ah@cybertec.at> Discussion: https://postgr.es/m/87lgdyz1wj.fsf@ars-thinkpad
2018-06-26 22:38:34 +02:00
txn = ReorderBufferTXNByXid(rb, txn->toplevel_xid, false,
NULL, InvalidXLogRecPtr, false);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
return txn->base_snapshot != NULL;
}
/*
* ---------------------------------------
* Disk serialization support
* ---------------------------------------
*/
/*
* Ensure the IO buffer is >= sz.
*/
static void
ReorderBufferSerializeReserve(ReorderBuffer *rb, Size sz)
{
if (!rb->outbufsize)
{
rb->outbuf = MemoryContextAlloc(rb->context, sz);
rb->outbufsize = sz;
}
else if (rb->outbufsize < sz)
{
rb->outbuf = repalloc(rb->outbuf, sz);
rb->outbufsize = sz;
}
}
/*
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
* Find the largest transaction (toplevel or subxact) to evict (spill to disk).
*
* XXX With many subtransactions this might be quite slow, because we'll have
* to walk through all of them. There are some options how we could improve
* that: (a) maintain some secondary structure with transactions sorted by
* amount of changes, (b) not looking for the entirely largest transaction,
* but e.g. for transaction using at least some fraction of the memory limit,
* and (c) evicting multiple transactions at once, e.g. to free a given portion
* of the memory limit (e.g. 50%).
*/
static ReorderBufferTXN *
ReorderBufferLargestTXN(ReorderBuffer *rb)
{
HASH_SEQ_STATUS hash_seq;
ReorderBufferTXNByIdEnt *ent;
ReorderBufferTXN *largest = NULL;
hash_seq_init(&hash_seq, rb->by_txn);
while ((ent = hash_seq_search(&hash_seq)) != NULL)
{
ReorderBufferTXN *txn = ent->txn;
/* if the current transaction is larger, remember it */
if ((!largest) || (txn->size > largest->size))
largest = txn;
}
Assert(largest);
Assert(largest->size > 0);
Assert(largest->size <= rb->size);
return largest;
}
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
/*
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
* Find the largest streamable toplevel transaction to evict (by streaming).
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
*
* This can be seen as an optimized version of ReorderBufferLargestTXN, which
* should give us the same transaction (because we don't update memory account
* for subtransaction with streaming, so it's always 0). But we can simply
Fix the bugs in selecting the transaction for streaming. There were two problems: a. We were always selecting the next available txn instead of selecting it when it is larger than the previous transaction. b. We were selecting the transactions which haven't made any changes to the database (base snapshot is not set). Later it was hitting an Assert because we don't decode such transactions and the changes in txn remain as it is. It is better not to choose such transactions for streaming in the first place. Reported-by: Haiying Tang Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/OS0PR01MB61133B94E63177040F7ECDA1FB429@OS0PR01MB6113.jpnprd01.prod.outlook.com
2021-04-30 07:19:52 +02:00
* iterate over the limited number of toplevel transactions that have a base
* snapshot. There is no use of selecting a transaction that doesn't have base
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
* snapshot because we don't decode such transactions. Also, we do not select
* the transaction which doesn't have any streamable change.
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
*
* Note that, we skip transactions that contains incomplete changes. There
Fix the bugs in selecting the transaction for streaming. There were two problems: a. We were always selecting the next available txn instead of selecting it when it is larger than the previous transaction. b. We were selecting the transactions which haven't made any changes to the database (base snapshot is not set). Later it was hitting an Assert because we don't decode such transactions and the changes in txn remain as it is. It is better not to choose such transactions for streaming in the first place. Reported-by: Haiying Tang Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/OS0PR01MB61133B94E63177040F7ECDA1FB429@OS0PR01MB6113.jpnprd01.prod.outlook.com
2021-04-30 07:19:52 +02:00
* is a scope of optimization here such that we can select the largest
* transaction which has incomplete changes. But that will make the code and
* design quite complex and that might not be worth the benefit. If we plan to
* stream the transactions that contains incomplete changes then we need to
* find a way to partially stream/truncate the transaction changes in-memory
* and build a mechanism to partially truncate the spilled files.
* Additionally, whenever we partially stream the transaction we need to
* maintain the last streamed lsn and next time we need to restore from that
* segment and the offset in WAL. As we stream the changes from the top
* transaction and restore them subtransaction wise, we need to even remember
* the subxact from where we streamed the last change.
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
*/
static ReorderBufferTXN *
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
ReorderBufferLargestStreamableTopTXN(ReorderBuffer *rb)
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
{
dlist_iter iter;
Size largest_size = 0;
ReorderBufferTXN *largest = NULL;
Fix the bugs in selecting the transaction for streaming. There were two problems: a. We were always selecting the next available txn instead of selecting it when it is larger than the previous transaction. b. We were selecting the transactions which haven't made any changes to the database (base snapshot is not set). Later it was hitting an Assert because we don't decode such transactions and the changes in txn remain as it is. It is better not to choose such transactions for streaming in the first place. Reported-by: Haiying Tang Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/OS0PR01MB61133B94E63177040F7ECDA1FB429@OS0PR01MB6113.jpnprd01.prod.outlook.com
2021-04-30 07:19:52 +02:00
/* Find the largest top-level transaction having a base snapshot. */
dlist_foreach(iter, &rb->txns_by_base_snapshot_lsn)
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
{
ReorderBufferTXN *txn;
Fix the bugs in selecting the transaction for streaming. There were two problems: a. We were always selecting the next available txn instead of selecting it when it is larger than the previous transaction. b. We were selecting the transactions which haven't made any changes to the database (base snapshot is not set). Later it was hitting an Assert because we don't decode such transactions and the changes in txn remain as it is. It is better not to choose such transactions for streaming in the first place. Reported-by: Haiying Tang Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/OS0PR01MB61133B94E63177040F7ECDA1FB429@OS0PR01MB6113.jpnprd01.prod.outlook.com
2021-04-30 07:19:52 +02:00
txn = dlist_container(ReorderBufferTXN, base_snapshot_node, iter.cur);
/* must not be a subtxn */
Assert(!rbtxn_is_known_subxact(txn));
/* base_snapshot must be set */
Assert(txn->base_snapshot != NULL);
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
Fix the bugs in selecting the transaction for streaming. There were two problems: a. We were always selecting the next available txn instead of selecting it when it is larger than the previous transaction. b. We were selecting the transactions which haven't made any changes to the database (base snapshot is not set). Later it was hitting an Assert because we don't decode such transactions and the changes in txn remain as it is. It is better not to choose such transactions for streaming in the first place. Reported-by: Haiying Tang Author: Dilip Kumar Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/OS0PR01MB61133B94E63177040F7ECDA1FB429@OS0PR01MB6113.jpnprd01.prod.outlook.com
2021-04-30 07:19:52 +02:00
if ((largest == NULL || txn->total_size > largest_size) &&
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
(txn->total_size > 0) && !(rbtxn_has_partial_change(txn)) &&
rbtxn_has_streamable_change(txn))
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
{
largest = txn;
largest_size = txn->total_size;
}
}
return largest;
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
* Check whether the logical_decoding_work_mem limit was reached, and if yes
Remove extra whitespace in comments atop ReorderBufferCheckMemoryLimit. Backpatch-through: 13, where it was introduced
2020-07-06 05:14:33 +02:00
* pick the largest (sub)transaction at-a-time to evict and spill its changes to
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
* disk or send to the output plugin until we reach under the memory limit.
*
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
* If logical_replication_mode is set to "immediate", stream or serialize the
* changes immediately.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +02:00
* XXX At this point we select the transactions until we reach under the memory
* limit, but we might also adapt a more elaborate eviction strategy - for example
* evicting enough transactions to free certain fraction (e.g. 50%) of the memory
* limit.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static void
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
ReorderBufferCheckMemoryLimit(ReorderBuffer *rb)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
ReorderBufferTXN *txn;
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
/*
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
* Bail out if logical_replication_mode is buffered and we haven't exceeded
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
* the memory limit.
*/
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
if (logical_replication_mode == LOGICAL_REP_MODE_BUFFERED &&
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
rb->size < logical_decoding_work_mem * 1024L)
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
return;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
* If logical_replication_mode is immediate, loop until there's no change.
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
* Otherwise, loop until we reach under the memory limit. One might think
* that just by evicting the largest (sub)transaction we will come under
* the memory limit based on assumption that the selected transaction is
* at least as large as the most recent change (which caused us to go over
* the memory limit). However, that is not true because a user can reduce
* the logical_decoding_work_mem to a smaller value before the most recent
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +02:00
* change.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
while (rb->size >= logical_decoding_work_mem * 1024L ||
Rename GUC logical_decoding_mode to logical_replication_mode. Rename the developer option 'logical_decoding_mode' to the more flexible name 'logical_replication_mode' because doing so will make it easier to extend this option in the future to help test other areas of logical replication. Currently, it is used on the publisher side to allow streaming or serializing each change in logical decoding. In the upcoming patch, we are planning to use it on the subscriber. On the subscriber, it will allow serializing the changes to file and notifies the parallel apply workers to read and apply them at the end of the transaction. We discussed exposing this parameter as a subscription option but it did not seem advisable since it is primarily used for testing/debugging and there is no other such parameter. We also discussed having separate GUCs for publisher and subscriber but for current testing/debugging requirements, one GUC is sufficient. Author: Hou Zhijie Reviewed-by: Peter Smith, Kuroda Hayato, Sawada Masahiko, Amit Kapila Discussion: https://postgr.es/m/CAD21AoAy2c=Mx=FTCs+EwUsf2kQL5MmU3N18X84k0EmCXntK4g@mail.gmail.com Discussion: https://postgr.es/m/CAA4eK1+wyN6zpaHUkCLorEWNx75MG0xhMwcFhvjqm2KURZEAGw@mail.gmail.com
2023-01-30 03:32:08 +01:00
(logical_replication_mode == LOGICAL_REP_MODE_IMMEDIATE &&
Add 'logical_decoding_mode' GUC. This enables streaming or serializing changes immediately in logical decoding. This parameter is intended to be used to test logical decoding and replication of large transactions for which otherwise we need to generate the changes till logical_decoding_work_mem is reached. This helps in reducing the timing of existing tests related to logical replication of in-progress transactions and will help in writing tests for for the upcoming feature for parallelly applying large in-progress transactions. Author: Shi yu Reviewed-by: Sawada Masahiko, Shveta Mallik, Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi Discussion: https://postgr.es/m/OSZPR01MB63104E7449DBE41932DB19F1FD1B9@OSZPR01MB6310.jpnprd01.prod.outlook.com
2022-12-26 04:28:16 +01:00
rb->size > 0))
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +02:00
{
/*
* Pick the largest transaction (or subtransaction) and evict it from
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
* memory by streaming, if possible. Otherwise, spill to disk.
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +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
if (ReorderBufferCanStartStreaming(rb) &&
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
(txn = ReorderBufferLargestStreamableTopTXN(rb)) != NULL)
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
{
/* we know there has to be one, because the size is not zero */
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
Assert(txn && rbtxn_is_toptxn(txn));
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
Assert(txn->total_size > 0);
Assert(rb->size >= txn->total_size);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
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
ReorderBufferStreamTXN(rb, txn);
}
else
{
/*
* Pick the largest transaction (or subtransaction) and evict it
* from memory by serializing it to disk.
*/
txn = ReorderBufferLargestTXN(rb);
/* we know there has to be one, because the size is not zero */
Assert(txn);
Assert(txn->size > 0);
Assert(rb->size >= txn->size);
ReorderBufferSerializeTXN(rb, txn);
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +02:00
/*
* After eviction, the transaction should have no entries in memory,
* and should use 0 bytes for changes.
*/
Assert(txn->size == 0);
Assert(txn->nentries_mem == 0);
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Fix ReorderBuffer memory overflow check. Commit cec2edfa78 introduced logical_decoding_work_mem to limit ReorderBuffer memory usage. We spill the changes once the memory occupied by changes exceeds logical_decoding_work_mem. There was an assumption in the code that by evicting the largest (sub)transaction we will come under the memory limit as the selected transaction will be at least as large as the most recent change (which caused us to go over the memory limit). However, that is not true because a user can reduce the logical_decoding_work_mem to a smaller value before the most recent change. We fix it by allowing to evict the transactions until we reach under the memory limit. Reported-by: Fujii Masao Author: Amit Kapila Reviewed-by: Fujii Masao Backpatch-through: 13, where it was introduced Discussion: https://postgr.es/m/2b7ba291-22e0-a187-d167-9e5309a3458d@oss.nttdata.com
2020-06-10 06:50:10 +02:00
/* We must be under the memory limit now. */
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Assert(rb->size < logical_decoding_work_mem * 1024L);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Spill data of a large transaction (and its subtransactions) to disk.
*/
static void
ReorderBufferSerializeTXN(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
dlist_iter subtxn_i;
dlist_mutable_iter change_i;
int fd = -1;
XLogSegNo curOpenSegNo = 0;
Size spilled = 0;
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
Size size = txn->size;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
elog(DEBUG2, "spill %u changes in XID %u to disk",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
(uint32) txn->nentries_mem, txn->xid);
/* do the same to all child TXs */
dlist_foreach(subtxn_i, &txn->subtxns)
{
ReorderBufferTXN *subtxn;
subtxn = dlist_container(ReorderBufferTXN, node, subtxn_i.cur);
ReorderBufferSerializeTXN(rb, subtxn);
}
/* serialize changestream */
dlist_foreach_modify(change_i, &txn->changes)
{
ReorderBufferChange *change;
change = dlist_container(ReorderBufferChange, node, change_i.cur);
/*
* store in segment in which it belongs by start lsn, don't split over
* multiple segments tho
*/
Make WAL segment size configurable at initdb time. For performance reasons a larger segment size than the default 16MB can be useful. A larger segment size has two main benefits: Firstly, in setups using archiving, it makes it easier to write scripts that can keep up with higher amounts of WAL, secondly, the WAL has to be written and synced to disk less frequently. But at the same time large segment size are disadvantageous for smaller databases. So far the segment size had to be configured at compile time, often making it unrealistic to choose one fitting to a particularly load. Therefore change it to a initdb time setting. This includes a breaking changes to the xlogreader.h API, which now requires the current segment size to be configured. For that and similar reasons a number of binaries had to be taught how to recognize the current segment size. Author: Beena Emerson, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
if (fd == -1 ||
!XLByteInSeg(change->lsn, curOpenSegNo, wal_segment_size))
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
char path[MAXPGPATH];
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (fd != -1)
CloseTransientFile(fd);
Make WAL segment size configurable at initdb time. For performance reasons a larger segment size than the default 16MB can be useful. A larger segment size has two main benefits: Firstly, in setups using archiving, it makes it easier to write scripts that can keep up with higher amounts of WAL, secondly, the WAL has to be written and synced to disk less frequently. But at the same time large segment size are disadvantageous for smaller databases. So far the segment size had to be configured at compile time, often making it unrealistic to choose one fitting to a particularly load. Therefore change it to a initdb time setting. This includes a breaking changes to the xlogreader.h API, which now requires the current segment size to be configured. For that and similar reasons a number of binaries had to be taught how to recognize the current segment size. Author: Beena Emerson, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
XLByteToSeg(change->lsn, curOpenSegNo, wal_segment_size);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* No need to care about TLIs here, only used during a single run,
* so each LSN only maps to a specific WAL record.
*/
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
ReorderBufferSerializedPath(path, MyReplicationSlot, txn->xid,
curOpenSegNo);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* open segment, create it if necessary */
fd = OpenTransientFile(path,
Refactor new file permission handling The file handling functions from fd.c were called with a diverse mix of notations for the file permissions when they were opening new files. Almost all files created by the server should have the same permissions set. So change the API so that e.g. OpenTransientFile() automatically uses the standard permissions set, and OpenTransientFilePerm() is a new function that takes an explicit permissions set for the few cases where it is needed. This also saves an unnecessary argument for call sites that are just opening an existing file. While we're reviewing these APIs, get rid of the FileName typedef and use the standard const char * for the file name and mode_t for the file mode. This makes these functions match other file handling functions and removes an unnecessary layer of mysteriousness. We can also get rid of a few casts that way. Author: David Steele <david@pgmasters.net>
2017-09-23 15:49:22 +02:00
O_CREAT | O_WRONLY | O_APPEND | PG_BINARY);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (fd < 0)
ereport(ERROR,
(errcode_for_file_access(),
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
errmsg("could not open file \"%s\": %m", path)));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
ReorderBufferSerializeChange(rb, txn, fd, change);
dlist_delete(&change->node);
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
spilled++;
}
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
/* update the statistics iff we have spilled anything */
if (spilled)
{
rb->spillCount += 1;
rb->spillBytes += size;
/* don't consider already serialized transactions */
rb->spillTxns += (rbtxn_is_serialized(txn) || rbtxn_is_serialized_clear(txn)) ? 0 : 1;
Update replication statistics after every stream/spill. Currently, replication slot statistics are updated at prepare, commit, and rollback. Now, if the transaction is interrupted the stats might not get updated. Fixed this by updating replication statistics after every stream/spill. In passing update the docs to change the description of some of the slot stats. Author: Vignesh C, Sawada Masahiko Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-06 07:51:26 +02:00
/* update the decoding stats */
UpdateDecodingStats((LogicalDecodingContext *) rb->private_data);
Track statistics for spilling of changes from ReorderBuffer. This adds the statistics about transactions spilled to disk from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. This commit extends the statistics collector to track this information about slots. Author: Sawada Masahiko and Amit Kapila Reviewed-by: Amit Kapila and Dilip Kumar Discussion: https://postgr.es/m/CA+fd4k5_pPAYRTDrO2PbtTOe0eHQpBvuqmCr8ic39uTNmR49Eg@mail.gmail.com
2020-10-08 05:39:08 +02:00
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(spilled == txn->nentries_mem);
Assert(dlist_is_empty(&txn->changes));
txn->nentries_mem = 0;
Clean up representation of flags in struct ReorderBufferTXN This simplifies addition of further flags. Author: Nikhil Sontakke Discussion: https://postgr.es/m/CAMGcDxeViP+R-OL7QhzUV9eKCVjURobuY1Zijik4Ay_Ddwo4Cg@mail.gmail.com
2020-01-10 21:46:57 +01:00
txn->txn_flags |= RBTXN_IS_SERIALIZED;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (fd != -1)
CloseTransientFile(fd);
}
/*
* Serialize individual change to disk.
*/
static void
ReorderBufferSerializeChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
int fd, ReorderBufferChange *change)
{
ReorderBufferDiskChange *ondisk;
Size sz = sizeof(ReorderBufferDiskChange);
ReorderBufferSerializeReserve(rb, sz);
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
memcpy(&ondisk->change, change, sizeof(ReorderBufferChange));
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
switch (change->action)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/* fall through these, they're all similar enough */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
case REORDER_BUFFER_CHANGE_INSERT:
case REORDER_BUFFER_CHANGE_UPDATE:
case REORDER_BUFFER_CHANGE_DELETE:
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT:
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
char *data;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferTupleBuf *oldtup,
*newtup;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Size oldlen = 0;
Size newlen = 0;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
oldtup = change->data.tp.oldtuple;
newtup = change->data.tp.newtuple;
if (oldtup)
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
{
sz += sizeof(HeapTupleData);
oldlen = oldtup->tuple.t_len;
sz += oldlen;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (newtup)
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
{
sz += sizeof(HeapTupleData);
newlen = newtup->tuple.t_len;
sz += newlen;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* make sure we have enough space */
ReorderBufferSerializeReserve(rb, sz);
data = ((char *) rb->outbuf) + sizeof(ReorderBufferDiskChange);
/* might have been reallocated above */
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
if (oldlen)
{
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
memcpy(data, &oldtup->tuple, sizeof(HeapTupleData));
data += sizeof(HeapTupleData);
memcpy(data, oldtup->tuple.t_data, oldlen);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
data += oldlen;
}
if (newlen)
{
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
memcpy(data, &newtup->tuple, sizeof(HeapTupleData));
data += sizeof(HeapTupleData);
memcpy(data, newtup->tuple.t_data, newlen);
Further improvements to c8f621c43. Coverity and inspection for the issue addressed in fd45d16f found some questionable code. Specifically coverity noticed that the wrong length was added in ReorderBufferSerializeChange() - without immediate negative consequences as the variable isn't used afterwards. During code-review and testing I noticed that a bit of space was wasted when allocating tuple bufs in several places. Thirdly, the debug memset()s in ReorderBufferGetTupleBuf() reduce the error checking valgrind can do. Backpatch: 9.4, like c8f621c43.
2016-03-07 23:24:03 +01:00
data += newlen;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
break;
}
case REORDER_BUFFER_CHANGE_MESSAGE:
{
char *data;
Size prefix_size = strlen(change->data.msg.prefix) + 1;
sz += prefix_size + change->data.msg.message_size +
sizeof(Size) + sizeof(Size);
ReorderBufferSerializeReserve(rb, sz);
data = ((char *) rb->outbuf) + sizeof(ReorderBufferDiskChange);
Fix dangling pointer problem in ReorderBufferSerializeChange. Commit 3fe3511d05127cc024b221040db2eeb352e7d716 introduced a new case into this function, but neglected to ensure that the "ondisk" pointer got updated after a possible reallocation as the code does in other cases. Stas Kelvich, per diagnosis by Konstantin Knizhnik.
2016-09-28 17:19:46 +02:00
/* might have been reallocated above */
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
/* write the prefix including the size */
memcpy(data, &prefix_size, sizeof(Size));
data += sizeof(Size);
memcpy(data, change->data.msg.prefix,
prefix_size);
data += prefix_size;
/* write the message including the size */
memcpy(data, &change->data.msg.message_size, sizeof(Size));
data += sizeof(Size);
memcpy(data, change->data.msg.message,
change->data.msg.message_size);
data += change->data.msg.message_size;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
break;
}
case REORDER_BUFFER_CHANGE_INVALIDATION:
{
char *data;
Size inval_size = sizeof(SharedInvalidationMessage) *
change->data.inval.ninvalidations;
sz += inval_size;
ReorderBufferSerializeReserve(rb, sz);
data = ((char *) rb->outbuf) + sizeof(ReorderBufferDiskChange);
/* might have been reallocated above */
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
memcpy(data, change->data.inval.invalidations, inval_size);
data += inval_size;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
break;
}
case REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT:
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Snapshot snap;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
char *data;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
snap = change->data.snapshot;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
sz += sizeof(SnapshotData) +
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
sizeof(TransactionId) * snap->xcnt +
Set ReorderBufferTXN->final_lsn more eagerly ... specifically, set it incrementally as each individual change is spilled down to disk. This way, it is set correctly when the transaction disappears without trace, ie. without leaving an XACT_ABORT wal record. (This happens when the server crashes midway through a transaction.) Failing to have final_lsn prevents ReorderBufferRestoreCleanup() from working, since it needs the final_lsn in order to know the endpoint of its iteration through spilled files. Commit df9f682c7bf8 already tried to fix the problem, but it didn't set the final_lsn in all cases. Revert that, since it's no longer needed. Author: Vignesh C Reviewed-by: Amit Kapila, Dilip Kumar Discussion: https://postgr.es/m/CALDaNm2CLk+K9JDwjYST0sPbGg5AQdvhUt0jbKyX_HdAE0jk3A@mail.gmail.com
2020-01-17 22:00:39 +01:00
sizeof(TransactionId) * snap->subxcnt;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* make sure we have enough space */
ReorderBufferSerializeReserve(rb, sz);
data = ((char *) rb->outbuf) + sizeof(ReorderBufferDiskChange);
/* might have been reallocated above */
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
memcpy(data, snap, sizeof(SnapshotData));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
data += sizeof(SnapshotData);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (snap->xcnt)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
memcpy(data, snap->xip,
Code review for logical decoding patch. Post-commit review identified a number of places where addition was used instead of multiplication or memory wasn't zeroed where it should have been. This commit also fixes one case where a structure member was mis-initialized, and moves another memory allocation closer to the place where the allocated storage is used for clarity. Andres Freund
2014-05-09 16:44:04 +02:00
sizeof(TransactionId) * snap->xcnt);
data += sizeof(TransactionId) * snap->xcnt;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
if (snap->subxcnt)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
memcpy(data, snap->subxip,
Code review for logical decoding patch. Post-commit review identified a number of places where addition was used instead of multiplication or memory wasn't zeroed where it should have been. This commit also fixes one case where a structure member was mis-initialized, and moves another memory allocation closer to the place where the allocated storage is used for clarity. Andres Freund
2014-05-09 16:44:04 +02:00
sizeof(TransactionId) * snap->subxcnt);
data += sizeof(TransactionId) * snap->subxcnt;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
break;
}
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
case REORDER_BUFFER_CHANGE_TRUNCATE:
Fix memory leak in TRUNCATE decoding When decoding a TRUNCATE record, the relids array was being allocated in the main ReorderBuffer memory context, but not released with the change resulting in a memory leak. The array was also ignored when serializing/deserializing the change, assuming all the information is stored in the change itself. So when spilling the change to disk, we've only we have serialized only the pointer to the relids array. Thanks to never releasing the array, the pointer however remained valid even after loading the change back to memory, preventing an actual crash. This fixes both the memory leak and (de)serialization. The relids array is still allocated in the main ReorderBuffer memory context (none of the existing ones seems like a good match, and adding an extra context seems like an overkill). The allocation is wrapped in a new ReorderBuffer API functions, to keep the details within reorderbuffer.c, just like the other ReorderBufferGet methods do. Author: Tomas Vondra Discussion: https://www.postgresql.org/message-id/flat/66175a41-9342-2845-652f-1bd4c3ee50aa%402ndquadrant.com Backpatch: 11, where decoding of TRUNCATE was introduced
2018-09-03 02:10:24 +02:00
{
Size size;
char *data;
/* account for the OIDs of truncated relations */
size = sizeof(Oid) * change->data.truncate.nrelids;
sz += size;
/* make sure we have enough space */
ReorderBufferSerializeReserve(rb, sz);
data = ((char *) rb->outbuf) + sizeof(ReorderBufferDiskChange);
/* might have been reallocated above */
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
memcpy(data, change->data.truncate.relids, size);
data += size;
break;
}
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM:
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT:
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID:
case REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID:
/* ReorderBufferChange contains everything important */
break;
}
ondisk->size = sz;
Reset properly errno before calling write() 6cb3372 enforces errno to ENOSPC when less bytes than what is expected have been written when it is unset, though it forgot to properly reset errno before doing a system call to write(), causing errno to potentially come from a previous system call. Reported-by: Tom Lane Author: Michael Paquier Reviewed-by: Tom Lane Discussion: https://postgr.es/m/31797.1533326676@sss.pgh.pa.us
2018-08-04 22:31:18 +02:00
errno = 0;
Create and use wait events for read, write, and fsync operations. Previous commits, notably 53be0b1add7064ca5db3cd884302dfc3268d884e and 6f3bd98ebfc008cbd676da777bb0b2376c4c4bfa, made it possible to see from pg_stat_activity when a backend was stuck waiting for another backend, but it's also fairly common for a backend to be stuck waiting for an I/O. Add wait events for those operations, too. Rushabh Lathia, with further hacking by me. Reviewed and tested by Michael Paquier, Amit Kapila, Rajkumar Raghuwanshi, and Rahila Syed. Discussion: http://postgr.es/m/CAGPqQf0LsYHXREPAZqYGVkDqHSyjf=KsD=k0GTVPAuzyThh-VQ@mail.gmail.com
2017-03-18 12:43:01 +01:00
pgstat_report_wait_start(WAIT_EVENT_REORDER_BUFFER_WRITE);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (write(fd, rb->outbuf, ondisk->size) != ondisk->size)
{
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
int save_errno = errno;
reorderbuffer: preserve errno while reporting error Clobbering errno during cleanup after an error is an oft-repeated, easy to make mistake. Deal with it here as everywhere else, by saving it aside and restoring after cleanup, before ereport'ing. In passing, add a missing errcode declaration in another ereport() call in the same file, which I noticed while skimming the file looking for similar problems. Backpatch to 9.4, where this code was introduced.
2016-08-19 19:38:55 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
CloseTransientFile(fd);
Address set of issues with errno handling System calls mixed up in error code paths are causing two issues which several code paths have not correctly handled: 1) For write() calls, sometimes the system may return less bytes than what has been written without errno being set. Some paths were careful enough to consider that case, and assumed that errno should be set to ENOSPC, other calls missed that. 2) errno generated by a system call is overwritten by other system calls which may succeed once an error code path is taken, causing what is reported to the user to be incorrect. This patch uses the brute-force approach of correcting all those code paths. Some refactoring could happen in the future, but this is let as future work, which is not targeted for back-branches anyway. Author: Michael Paquier Reviewed-by: Ashutosh Sharma Discussion: https://postgr.es/m/20180622061535.GD5215@paquier.xyz
2018-06-25 04:19:05 +02:00
/* if write didn't set errno, assume problem is no disk space */
errno = save_errno ? save_errno : ENOSPC;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ereport(ERROR,
(errcode_for_file_access(),
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
errmsg("could not write to data file for XID %u: %m",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
txn->xid)));
}
Create and use wait events for read, write, and fsync operations. Previous commits, notably 53be0b1add7064ca5db3cd884302dfc3268d884e and 6f3bd98ebfc008cbd676da777bb0b2376c4c4bfa, made it possible to see from pg_stat_activity when a backend was stuck waiting for another backend, but it's also fairly common for a backend to be stuck waiting for an I/O. Add wait events for those operations, too. Rushabh Lathia, with further hacking by me. Reviewed and tested by Michael Paquier, Amit Kapila, Rajkumar Raghuwanshi, and Rahila Syed. Discussion: http://postgr.es/m/CAGPqQf0LsYHXREPAZqYGVkDqHSyjf=KsD=k0GTVPAuzyThh-VQ@mail.gmail.com
2017-03-18 12:43:01 +01:00
pgstat_report_wait_end();
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Set ReorderBufferTXN->final_lsn more eagerly ... specifically, set it incrementally as each individual change is spilled down to disk. This way, it is set correctly when the transaction disappears without trace, ie. without leaving an XACT_ABORT wal record. (This happens when the server crashes midway through a transaction.) Failing to have final_lsn prevents ReorderBufferRestoreCleanup() from working, since it needs the final_lsn in order to know the endpoint of its iteration through spilled files. Commit df9f682c7bf8 already tried to fix the problem, but it didn't set the final_lsn in all cases. Revert that, since it's no longer needed. Author: Vignesh C Reviewed-by: Amit Kapila, Dilip Kumar Discussion: https://postgr.es/m/CALDaNm2CLk+K9JDwjYST0sPbGg5AQdvhUt0jbKyX_HdAE0jk3A@mail.gmail.com
2020-01-17 22:00:39 +01:00
/*
* Keep the transaction's final_lsn up to date with each change we send to
* disk, so that ReorderBufferRestoreCleanup works correctly. (We used to
* only do this on commit and abort records, but that doesn't work if a
* system crash leaves a transaction without its abort record).
*
* Make sure not to move it backwards.
*/
if (txn->final_lsn < change->lsn)
txn->final_lsn = change->lsn;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(ondisk->change.action == change->action);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
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
/* Returns true, if the output plugin supports streaming, false, otherwise. */
static inline bool
ReorderBufferCanStream(ReorderBuffer *rb)
{
LogicalDecodingContext *ctx = rb->private_data;
return ctx->streaming;
}
/* Returns true, if the streaming can be started now, false, otherwise. */
static inline bool
ReorderBufferCanStartStreaming(ReorderBuffer *rb)
{
LogicalDecodingContext *ctx = rb->private_data;
SnapBuild *builder = ctx->snapshot_builder;
Remove incorrect assertion in reorderbuffer.c. We start recording changes in ReorderBufferTXN even before we reach SNAPBUILD_CONSISTENT state so that if the commit is encountered after reaching that we should be able to send the changes of the entire transaction. Now, while recording changes if the reorder buffer memory has exceeded logical_decoding_work_mem then we can start streaming if it is allowed and we haven't yet streamed that data. However, we must not allow streaming to start unless the snapshot has reached SNAPBUILD_CONSISTENT state. In passing, improve the comments atop ReorderBufferResetTXN to mention the case when we need to continue streaming after getting an error. Author: Amit Kapila Reviewed-by: Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1KoOH0byboyYY40NBcC7Fe812trwTa+WY3jQF7WQWZbQg@mail.gmail.com
2020-12-04 09:24:50 +01:00
/* We can't start streaming unless a consistent state is reached. */
if (SnapBuildCurrentState(builder) < SNAPBUILD_CONSISTENT)
return false;
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
/*
* We can't start streaming immediately even if the streaming is enabled
* because we previously decoded this transaction and now just are
* restarting.
*/
if (ReorderBufferCanStream(rb) &&
Avoid unnecessary streaming of transactions during logical replication. After restart, we don't perform streaming of an in-progress transaction if it was previously decoded and confirmed by the client. To achieve that we were comparing the END location of the WAL record being decoded with the WAL location we have already decoded and confirmed by the client. While decoding the commit record, to decide whether to process and send the complete transaction, we compare its START location with the WAL location we have already decoded and confirmed by the client. Now, if we need to queue some change in the transaction while decoding the commit record (e.g. snapshot), it is possible that we decide to stream the transaction but later commit processing decides to skip it. In such a case, we would needlessly send the changes and later when we decide to skip it, we will send stream abort. We also sometimes decide to stream the changes when we actually just need to process them locally like a change for invalidations. This will lead us to send empty streams. To avoid this, while queuing each change for decoding, we remember whether the transaction has any change that actually needs to be sent downstream and use that information later to decide whether to stream the transaction or not. Note, we can't avoid all cases where we have to send empty streams like the case where the plugin later decides that the change is not publishable. However, we will no longer need to send stream_abort when we skip sending a particular transaction. Author: Dilip Kumar Reviewed-by: Hou Zhijie, Ashutosh Bapat, Shi yu, Amit Kapila Discussion: https://postgr.es/m/CAFiTN-tHK=7LzfrPs8fbT2ksrOJGQbzywcgXst2bM9-rJJAAUg@mail.gmail.com
2022-12-08 01:35:09 +01:00
!SnapBuildXactNeedsSkip(builder, ctx->reader->ReadRecPtr))
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
return true;
return false;
}
/*
* Send data of a large transaction (and its subtransactions) to the
* output plugin, but using the stream API.
*/
static void
ReorderBufferStreamTXN(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
Snapshot snapshot_now;
CommandId command_id;
Track statistics for streaming of changes from ReorderBuffer. This adds the statistics about transactions streamed to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. Commit 9868167500 has added the basic infrastructure to capture the stats of slot and this commit extends the statistics collector to track additional information about slots. Bump the catversion as we have added new columns in the catalog entry. Author: Ajin Cherian and Amit Kapila Reviewed-by: Sawada Masahiko and Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1+chpEomLzgSoky-D31qev19AmECNiEAietPQUGEFhtVA@mail.gmail.com
2020-10-29 04:41:51 +01:00
Size stream_bytes;
bool txn_is_streamed;
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
/* We can never reach here for a subtransaction. */
Add macros for ReorderBufferTXN toptxn. Currently, there are quite a few places in reorderbuffer.c that tries to access top-transaction for a subtransaction. This makes the code to access top-transaction consistent and easier to follow. Author: Peter Smith Reviewed-by: Vignesh C, Sawada Masahiko Discussion: https://postgr.es/m/CAHut+PuCznOyTqBQwjRUu-ibG-=KHyCv-0FTcWQtZUdR88umfg@mail.gmail.com
2023-03-17 03:59:41 +01:00
Assert(rbtxn_is_toptxn(txn));
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
/*
* We can't make any assumptions about base snapshot here, similar to what
* ReorderBufferCommit() does. That relies on base_snapshot getting
* transferred from subxact in ReorderBufferCommitChild(), but that was
* not yet called as the transaction is in-progress.
*
* So just walk the subxacts and use the same logic here. But we only need
* to do that once, when the transaction is streamed for the first time.
* After that we need to reuse the snapshot from the previous run.
*
* Unlike DecodeCommit which adds xids of all the subtransactions in
Fix typos in comments The changes done in this commit impact comments with no direct user-visible changes, with fixes for incorrect function, variable or structure names. Author: Alexander Lakhin Discussion: https://postgr.es/m/e8c38840-596a-83d6-bd8d-cebc51111572@gmail.com
2023-05-02 05:23:08 +02:00
* snapshot's xip array via SnapBuildCommitTxn, we can't do that here
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
* but we do add them to subxip array instead via ReorderBufferCopySnap.
* This allows the catalog changes made in subtransactions decoded till
* now to be visible.
*/
if (txn->snapshot_now == NULL)
{
dlist_iter subxact_i;
/* make sure this transaction is streamed for the first time */
Assert(!rbtxn_is_streamed(txn));
/* at the beginning we should have invalid command ID */
Assert(txn->command_id == InvalidCommandId);
dlist_foreach(subxact_i, &txn->subtxns)
{
ReorderBufferTXN *subtxn;
subtxn = dlist_container(ReorderBufferTXN, node, subxact_i.cur);
ReorderBufferTransferSnapToParent(txn, subtxn);
}
/*
* If this transaction has no snapshot, it didn't make any changes to
* the database till now, so there's nothing to decode.
*/
if (txn->base_snapshot == NULL)
{
Assert(txn->ninvalidations == 0);
return;
}
command_id = FirstCommandId;
snapshot_now = ReorderBufferCopySnap(rb, txn->base_snapshot,
txn, command_id);
}
else
{
/* the transaction must have been already streamed */
Assert(rbtxn_is_streamed(txn));
/*
* Nah, we already have snapshot from the previous streaming run. We
* assume new subxacts can't move the LSN backwards, and so can't beat
* the LSN condition in the previous branch (so no need to walk
* through subxacts again). In fact, we must not do that as we may be
* using snapshot half-way through the subxact.
*/
command_id = txn->command_id;
/*
* We can't use txn->snapshot_now directly because after the last
* streaming run, we might have got some new sub-transactions. So we
* need to add them to the snapshot.
*/
snapshot_now = ReorderBufferCopySnap(rb, txn->snapshot_now,
txn, command_id);
/* Free the previously copied snapshot. */
Assert(txn->snapshot_now->copied);
ReorderBufferFreeSnap(rb, txn->snapshot_now);
txn->snapshot_now = NULL;
}
Track statistics for streaming of changes from ReorderBuffer. This adds the statistics about transactions streamed to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. Commit 9868167500 has added the basic infrastructure to capture the stats of slot and this commit extends the statistics collector to track additional information about slots. Bump the catversion as we have added new columns in the catalog entry. Author: Ajin Cherian and Amit Kapila Reviewed-by: Sawada Masahiko and Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1+chpEomLzgSoky-D31qev19AmECNiEAietPQUGEFhtVA@mail.gmail.com
2020-10-29 04:41:51 +01:00
/*
* Remember this information to be used later to update stats. We can't
* update the stats here as an error while processing the changes would
* lead to the accumulation of stats even though we haven't streamed all
* the changes.
*/
txn_is_streamed = rbtxn_is_streamed(txn);
stream_bytes = txn->total_size;
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
/* Process and send the changes to output plugin. */
ReorderBufferProcessTXN(rb, txn, InvalidXLogRecPtr, snapshot_now,
command_id, true);
Track statistics for streaming of changes from ReorderBuffer. This adds the statistics about transactions streamed to the decoding output plugin from ReorderBuffer. Users can query the pg_stat_replication_slots view to check these stats and call pg_stat_reset_replication_slot to reset the stats of a particular slot. Users can pass NULL in pg_stat_reset_replication_slot to reset stats of all the slots. Commit 9868167500 has added the basic infrastructure to capture the stats of slot and this commit extends the statistics collector to track additional information about slots. Bump the catversion as we have added new columns in the catalog entry. Author: Ajin Cherian and Amit Kapila Reviewed-by: Sawada Masahiko and Dilip Kumar Discussion: https://postgr.es/m/CAA4eK1+chpEomLzgSoky-D31qev19AmECNiEAietPQUGEFhtVA@mail.gmail.com
2020-10-29 04:41:51 +01:00
rb->streamCount += 1;
rb->streamBytes += stream_bytes;
/* Don't consider already streamed transaction. */
rb->streamTxns += (txn_is_streamed) ? 0 : 1;
Update replication statistics after every stream/spill. Currently, replication slot statistics are updated at prepare, commit, and rollback. Now, if the transaction is interrupted the stats might not get updated. Fixed this by updating replication statistics after every stream/spill. In passing update the docs to change the description of some of the slot stats. Author: Vignesh C, Sawada Masahiko Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/20210319185247.ldebgpdaxsowiflw@alap3.anarazel.de
2021-05-06 07:51:26 +02:00
/* update the decoding stats */
UpdateDecodingStats((LogicalDecodingContext *) rb->private_data);
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
Assert(dlist_is_empty(&txn->changes));
Assert(txn->nentries == 0);
Assert(txn->nentries_mem == 0);
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
* Size of a change in memory.
*/
static Size
ReorderBufferChangeSize(ReorderBufferChange *change)
{
Size sz = sizeof(ReorderBufferChange);
switch (change->action)
{
/* fall through these, they're all similar enough */
case REORDER_BUFFER_CHANGE_INSERT:
case REORDER_BUFFER_CHANGE_UPDATE:
case REORDER_BUFFER_CHANGE_DELETE:
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT:
{
ReorderBufferTupleBuf *oldtup,
*newtup;
Size oldlen = 0;
Size newlen = 0;
oldtup = change->data.tp.oldtuple;
newtup = change->data.tp.newtuple;
if (oldtup)
{
sz += sizeof(HeapTupleData);
oldlen = oldtup->tuple.t_len;
sz += oldlen;
}
if (newtup)
{
sz += sizeof(HeapTupleData);
newlen = newtup->tuple.t_len;
sz += newlen;
}
break;
}
case REORDER_BUFFER_CHANGE_MESSAGE:
{
Size prefix_size = strlen(change->data.msg.prefix) + 1;
sz += prefix_size + change->data.msg.message_size +
sizeof(Size) + sizeof(Size);
break;
}
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
case REORDER_BUFFER_CHANGE_INVALIDATION:
{
sz += sizeof(SharedInvalidationMessage) *
change->data.inval.ninvalidations;
break;
}
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT:
{
Snapshot snap;
snap = change->data.snapshot;
sz += sizeof(SnapshotData) +
sizeof(TransactionId) * snap->xcnt +
sizeof(TransactionId) * snap->subxcnt;
break;
}
case REORDER_BUFFER_CHANGE_TRUNCATE:
{
sz += sizeof(Oid) * change->data.truncate.nrelids;
break;
}
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM:
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT:
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID:
case REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID:
/* ReorderBufferChange contains everything important */
break;
}
return sz;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Restore a number of changes spilled to disk back into memory.
*/
static Size
ReorderBufferRestoreChanges(ReorderBuffer *rb, ReorderBufferTXN *txn,
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
TXNEntryFile *file, XLogSegNo *segno)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Size restored = 0;
XLogSegNo last_segno;
dlist_mutable_iter cleanup_iter;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
File *fd = &file->vfd;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(txn->first_lsn != InvalidXLogRecPtr);
Assert(txn->final_lsn != InvalidXLogRecPtr);
/* free current entries, so we have memory for more */
dlist_foreach_modify(cleanup_iter, &txn->changes)
{
ReorderBufferChange *cleanup =
dlist_container(ReorderBufferChange, node, cleanup_iter.cur);
dlist_delete(&cleanup->node);
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
ReorderBufferReturnChange(rb, cleanup, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
txn->nentries_mem = 0;
Assert(dlist_is_empty(&txn->changes));
Make WAL segment size configurable at initdb time. For performance reasons a larger segment size than the default 16MB can be useful. A larger segment size has two main benefits: Firstly, in setups using archiving, it makes it easier to write scripts that can keep up with higher amounts of WAL, secondly, the WAL has to be written and synced to disk less frequently. But at the same time large segment size are disadvantageous for smaller databases. So far the segment size had to be configured at compile time, often making it unrealistic to choose one fitting to a particularly load. Therefore change it to a initdb time setting. This includes a breaking changes to the xlogreader.h API, which now requires the current segment size to be configured. For that and similar reasons a number of binaries had to be taught how to recognize the current segment size. Author: Beena Emerson, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
XLByteToSeg(txn->final_lsn, last_segno, wal_segment_size);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
while (restored < max_changes_in_memory && *segno <= last_segno)
{
int readBytes;
ReorderBufferDiskChange *ondisk;
Add CHECK_FOR_INTERRUPTS while restoring changes during decoding. Previously in commit 42681dffaf, we added CFI during decoding changes but missed another similar case that can happen while restoring changes spilled to disk back into memory in a loop. Reported-by: Robert Haas Author: Amit Kapila Backpatch-through: 10 Discussion: https://postgr.es/m/CA+TgmoaLObg0QbstbC8ykDwOdD1bDkr4AbPpB=0DPgA2JW0mFg@mail.gmail.com
2022-10-21 09:27:18 +02:00
CHECK_FOR_INTERRUPTS();
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (*fd == -1)
{
char path[MAXPGPATH];
/* first time in */
if (*segno == 0)
Make WAL segment size configurable at initdb time. For performance reasons a larger segment size than the default 16MB can be useful. A larger segment size has two main benefits: Firstly, in setups using archiving, it makes it easier to write scripts that can keep up with higher amounts of WAL, secondly, the WAL has to be written and synced to disk less frequently. But at the same time large segment size are disadvantageous for smaller databases. So far the segment size had to be configured at compile time, often making it unrealistic to choose one fitting to a particularly load. Therefore change it to a initdb time setting. This includes a breaking changes to the xlogreader.h API, which now requires the current segment size to be configured. For that and similar reasons a number of binaries had to be taught how to recognize the current segment size. Author: Beena Emerson, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
XLByteToSeg(txn->first_lsn, *segno, wal_segment_size);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(*segno != 0 || dlist_is_empty(&txn->changes));
/*
* No need to care about TLIs here, only used during a single run,
* so each LSN only maps to a specific WAL record.
*/
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
ReorderBufferSerializedPath(path, MyReplicationSlot, txn->xid,
*segno);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
*fd = PathNameOpenFile(path, O_RDONLY | PG_BINARY);
/* No harm in resetting the offset even in case of failure */
file->curOffset = 0;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (*fd < 0 && errno == ENOENT)
{
*fd = -1;
(*segno)++;
continue;
}
else if (*fd < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\": %m",
path)));
}
/*
* Read the statically sized part of a change which has information
* about the total size. If we couldn't read a record, we're at the
* end of this file.
*/
Code review for logical decoding patch. Post-commit review identified a number of places where addition was used instead of multiplication or memory wasn't zeroed where it should have been. This commit also fixes one case where a structure member was mis-initialized, and moves another memory allocation closer to the place where the allocated storage is used for clarity. Andres Freund
2014-05-09 16:44:04 +02:00
ReorderBufferSerializeReserve(rb, sizeof(ReorderBufferDiskChange));
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
readBytes = FileRead(file->vfd, rb->outbuf,
sizeof(ReorderBufferDiskChange),
file->curOffset, WAIT_EVENT_REORDER_BUFFER_READ);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* eof */
if (readBytes == 0)
{
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
FileClose(*fd);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*fd = -1;
(*segno)++;
continue;
}
else if (readBytes < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not read from reorderbuffer spill file: %m")));
else if (readBytes != sizeof(ReorderBufferDiskChange))
ereport(ERROR,
(errcode_for_file_access(),
Message improvements
2014-11-12 02:00:58 +01:00
errmsg("could not read from reorderbuffer spill file: read %d instead of %u bytes",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
readBytes,
(uint32) sizeof(ReorderBufferDiskChange))));
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
file->curOffset += readBytes;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
ReorderBufferSerializeReserve(rb,
sizeof(ReorderBufferDiskChange) + ondisk->size);
ondisk = (ReorderBufferDiskChange *) rb->outbuf;
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
readBytes = FileRead(file->vfd,
rb->outbuf + sizeof(ReorderBufferDiskChange),
ondisk->size - sizeof(ReorderBufferDiskChange),
file->curOffset,
WAIT_EVENT_REORDER_BUFFER_READ);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (readBytes < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not read from reorderbuffer spill file: %m")));
else if (readBytes != ondisk->size - sizeof(ReorderBufferDiskChange))
ereport(ERROR,
(errcode_for_file_access(),
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
errmsg("could not read from reorderbuffer spill file: read %d instead of %u bytes",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
readBytes,
(uint32) (ondisk->size - sizeof(ReorderBufferDiskChange)))));
Fix running out of file descriptors for spill files. Currently while decoding changes, if the number of changes exceeds a certain threshold, we spill those to disk.  And this happens for each (sub)transaction.  Now, while reading all these files, we don't close them until we read all the files.  While reading these files, if the number of such files exceeds the maximum number of file descriptors, the operation errors out. Use PathNameOpenFile interface to open these files as that internally has the mechanism to release kernel FDs as needed to get us under the max_safe_fds limit. Reported-by: Amit Khandekar Author: Amit Khandekar Reviewed-by: Amit Kapila Backpatch-through: 9.4 Discussion: https://postgr.es/m/CAJ3gD9c-sECEn79zXw4yBnBdOttacoE-6gAyP0oy60nfs_sabQ@mail.gmail.com
2019-12-14 07:11:37 +01:00
file->curOffset += readBytes;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* ok, read a full change from disk, now restore it into proper
* in-memory format
*/
ReorderBufferRestoreChange(rb, txn, rb->outbuf);
restored++;
}
return restored;
}
/*
* Convert change from its on-disk format to in-memory format and queue it onto
* the TXN's ->changes list.
Fix core dump in ReorderBufferRestoreChange on alignment-picky platforms. When re-reading an update involving both an old tuple and a new tuple from disk, reorderbuffer.c was careless about whether the new tuple is suitably aligned for direct access --- in general, it isn't. We'd missed seeing this in the buildfarm because the contrib/test_decoding tests exercise this code path only a few times, and by chance all of those cases have old tuples with length a multiple of 4, which is usually enough to make the access to the new tuple's t_len safe. For some still-not-entirely-clear reason, however, Debian's sparc build gets a bus error, as reported by Christoph Berg; perhaps it's assuming 8-byte alignment of the pointer? The lack of previous field reports is probably because you need all of these conditions to trigger a crash: an alignment-picky platform (not Intel), a transaction large enough to spill to disk, an update within that xact that changes a primary-key field and has an odd-length old tuple, and of course logical decoding tracing the transaction. Avoid the alignment assumption by using memcpy instead of fetching t_len directly, and add a test case that exposes the crash on picky platforms. Back-patch to 9.4 where the bug was introduced. Discussion: <20160413094117.GC21485@msg.credativ.de>
2016-04-15 01:42:21 +02:00
*
* Note: although "data" is declared char*, at entry it points to a
* maxalign'd buffer, making it safe in most of this function to assume
* that the pointed-to data is suitably aligned for direct access.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static void
ReorderBufferRestoreChange(ReorderBuffer *rb, ReorderBufferTXN *txn,
char *data)
{
ReorderBufferDiskChange *ondisk;
ReorderBufferChange *change;
ondisk = (ReorderBufferDiskChange *) data;
change = ReorderBufferGetChange(rb);
/* copy static part */
memcpy(change, &ondisk->change, sizeof(ReorderBufferChange));
data += sizeof(ReorderBufferDiskChange);
/* restore individual stuff */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
switch (change->action)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
/* fall through these, they're all similar enough */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
case REORDER_BUFFER_CHANGE_INSERT:
case REORDER_BUFFER_CHANGE_UPDATE:
case REORDER_BUFFER_CHANGE_DELETE:
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_INSERT:
logical decoding: old/newtuple in spooled UPDATE changes was switched around. Somehow I managed to flip the order of restoring old & new tuples when de-spooling a change in a large transaction from disk. This happens to only take effect when a change is spooled to disk which has old/new versions of the tuple. That only is the case for UPDATEs where he primary key changed or where replica identity is changed to FULL. The tests didn't catch this because either spooled updates, or updates that changed primary keys, were tested; not both at the same time. Found while adding tests for the following commit. Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
if (change->data.tp.oldtuple)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Fix core dump in ReorderBufferRestoreChange on alignment-picky platforms. When re-reading an update involving both an old tuple and a new tuple from disk, reorderbuffer.c was careless about whether the new tuple is suitably aligned for direct access --- in general, it isn't. We'd missed seeing this in the buildfarm because the contrib/test_decoding tests exercise this code path only a few times, and by chance all of those cases have old tuples with length a multiple of 4, which is usually enough to make the access to the new tuple's t_len safe. For some still-not-entirely-clear reason, however, Debian's sparc build gets a bus error, as reported by Christoph Berg; perhaps it's assuming 8-byte alignment of the pointer? The lack of previous field reports is probably because you need all of these conditions to trigger a crash: an alignment-picky platform (not Intel), a transaction large enough to spill to disk, an update within that xact that changes a primary-key field and has an odd-length old tuple, and of course logical decoding tracing the transaction. Avoid the alignment assumption by using memcpy instead of fetching t_len directly, and add a test case that exposes the crash on picky platforms. Back-patch to 9.4 where the bug was introduced. Discussion: <20160413094117.GC21485@msg.credativ.de>
2016-04-15 01:42:21 +02:00
uint32 tuplelen = ((HeapTuple) data)->t_len;
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
change->data.tp.oldtuple =
ReorderBufferGetTupleBuf(rb, tuplelen - SizeofHeapTupleHeader);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
/* restore ->tuple */
memcpy(&change->data.tp.oldtuple->tuple, data,
sizeof(HeapTupleData));
data += sizeof(HeapTupleData);
/* reset t_data pointer into the new tuplebuf */
logical decoding: old/newtuple in spooled UPDATE changes was switched around. Somehow I managed to flip the order of restoring old & new tuples when de-spooling a change in a large transaction from disk. This happens to only take effect when a change is spooled to disk which has old/new versions of the tuple. That only is the case for UPDATEs where he primary key changed or where replica identity is changed to FULL. The tests didn't catch this because either spooled updates, or updates that changed primary keys, were tested; not both at the same time. Found while adding tests for the following commit. Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
change->data.tp.oldtuple->tuple.t_data =
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
ReorderBufferTupleBufData(change->data.tp.oldtuple);
/* restore tuple data itself */
memcpy(change->data.tp.oldtuple->tuple.t_data, data, tuplelen);
data += tuplelen;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
logical decoding: old/newtuple in spooled UPDATE changes was switched around. Somehow I managed to flip the order of restoring old & new tuples when de-spooling a change in a large transaction from disk. This happens to only take effect when a change is spooled to disk which has old/new versions of the tuple. That only is the case for UPDATEs where he primary key changed or where replica identity is changed to FULL. The tests didn't catch this because either spooled updates, or updates that changed primary keys, were tested; not both at the same time. Found while adding tests for the following commit. Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
if (change->data.tp.newtuple)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Fix core dump in ReorderBufferRestoreChange on alignment-picky platforms. When re-reading an update involving both an old tuple and a new tuple from disk, reorderbuffer.c was careless about whether the new tuple is suitably aligned for direct access --- in general, it isn't. We'd missed seeing this in the buildfarm because the contrib/test_decoding tests exercise this code path only a few times, and by chance all of those cases have old tuples with length a multiple of 4, which is usually enough to make the access to the new tuple's t_len safe. For some still-not-entirely-clear reason, however, Debian's sparc build gets a bus error, as reported by Christoph Berg; perhaps it's assuming 8-byte alignment of the pointer? The lack of previous field reports is probably because you need all of these conditions to trigger a crash: an alignment-picky platform (not Intel), a transaction large enough to spill to disk, an update within that xact that changes a primary-key field and has an odd-length old tuple, and of course logical decoding tracing the transaction. Avoid the alignment assumption by using memcpy instead of fetching t_len directly, and add a test case that exposes the crash on picky platforms. Back-patch to 9.4 where the bug was introduced. Discussion: <20160413094117.GC21485@msg.credativ.de>
2016-04-15 01:42:21 +02:00
/* here, data might not be suitably aligned! */
uint32 tuplelen;
memcpy(&tuplelen, data + offsetof(HeapTupleData, t_len),
sizeof(uint32));
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
change->data.tp.newtuple =
ReorderBufferGetTupleBuf(rb, tuplelen - SizeofHeapTupleHeader);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
/* restore ->tuple */
memcpy(&change->data.tp.newtuple->tuple, data,
sizeof(HeapTupleData));
data += sizeof(HeapTupleData);
/* reset t_data pointer into the new tuplebuf */
logical decoding: old/newtuple in spooled UPDATE changes was switched around. Somehow I managed to flip the order of restoring old & new tuples when de-spooling a change in a large transaction from disk. This happens to only take effect when a change is spooled to disk which has old/new versions of the tuple. That only is the case for UPDATEs where he primary key changed or where replica identity is changed to FULL. The tests didn't catch this because either spooled updates, or updates that changed primary keys, were tested; not both at the same time. Found while adding tests for the following commit. Backpatch: 9.4, where logical decoding was added
2016-03-06 03:02:20 +01:00
change->data.tp.newtuple->tuple.t_data =
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
ReorderBufferTupleBufData(change->data.tp.newtuple);
/* restore tuple data itself */
memcpy(change->data.tp.newtuple->tuple.t_data, data, tuplelen);
data += tuplelen;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
break;
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
case REORDER_BUFFER_CHANGE_MESSAGE:
{
Size prefix_size;
/* read prefix */
memcpy(&prefix_size, data, sizeof(Size));
data += sizeof(Size);
change->data.msg.prefix = MemoryContextAlloc(rb->context,
prefix_size);
memcpy(change->data.msg.prefix, data, prefix_size);
Assert(change->data.msg.prefix[prefix_size - 1] == '\0');
data += prefix_size;
Fix typos in comments. Vinayak Pokale
2016-10-26 10:12:31 +02:00
/* read the message */
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
memcpy(&change->data.msg.message_size, data, sizeof(Size));
data += sizeof(Size);
change->data.msg.message = MemoryContextAlloc(rb->context,
change->data.msg.message_size);
memcpy(change->data.msg.message, data,
change->data.msg.message_size);
data += change->data.msg.message_size;
Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message during logical decoding. Prior to commit c55040ccd0 we have no way of knowing the invalidations before commit. So, while decoding we use to execute all the invalidations at each command end as we had no way of knowing which invalidations happened before that command. Due to this, transactions involving large amounts of DDLs use to take more time and also lead to high CPU usage. But now we know specific invalidations at each command end so we execute only required invalidations. It has been observed that decoding of a transaction containing truncation of a table with 1000 partitions would be finished in 1s whereas before this patch it used to take 4-5 minutes. Author: Dilip Kumar Reviewed-by: Amit Kapila and Keisuke Kuroda Discussion: https://postgr.es/m/CANDwggKYveEtXjXjqHA6RL3AKSHMsQyfRY6bK+NqhAWJyw8psQ@mail.gmail.com
2020-10-15 04:47:51 +02:00
break;
}
case REORDER_BUFFER_CHANGE_INVALIDATION:
{
Size inval_size = sizeof(SharedInvalidationMessage) *
change->data.inval.ninvalidations;
change->data.inval.invalidations =
MemoryContextAlloc(rb->context, inval_size);
/* read the message */
memcpy(change->data.inval.invalidations, data, inval_size);
Generic Messages for Logical Decoding API and mechanism to allow generic messages to be inserted into WAL that are intended to be read by logical decoding plugins. This commit adds an optional new callback to the logical decoding API. Messages are either text or bytea. Messages can be transactional, or not, and are identified by a prefix to allow multiple concurrent decoding plugins. (Not to be confused with Generic WAL records, which are intended to allow crash recovery of extensible objects.) Author: Petr Jelinek and Andres Freund Reviewers: Artur Zakirov, Tomas Vondra, Simon Riggs Discussion: 5685F999.6010202@2ndquadrant.com
2016-04-06 11:05:41 +02:00
break;
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_SNAPSHOT:
{
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Snapshot oldsnap;
Snapshot newsnap;
Size size;
oldsnap = (Snapshot) data;
size = sizeof(SnapshotData) +
sizeof(TransactionId) * oldsnap->xcnt +
sizeof(TransactionId) * (oldsnap->subxcnt + 0);
change->data.snapshot = MemoryContextAllocZero(rb->context, size);
newsnap = change->data.snapshot;
memcpy(newsnap, data, size);
newsnap->xip = (TransactionId *)
(((char *) newsnap) + sizeof(SnapshotData));
newsnap->subxip = newsnap->xip + newsnap->xcnt;
newsnap->copied = true;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
break;
}
/* the base struct contains all the data, easy peasy */
Logical decoding of TRUNCATE Add a new WAL record type for TRUNCATE, which is only used when wal_level >= logical. (For physical replication, TRUNCATE is already replicated via SMGR records.) Add new callback for logical decoding output plugins to receive TRUNCATE actions. Author: Simon Riggs <simon@2ndquadrant.com> Author: Marco Nenciarini <marco.nenciarini@2ndquadrant.it> Author: Peter Eisentraut <peter.eisentraut@2ndquadrant.com> Reviewed-by: Petr Jelinek <petr.jelinek@2ndquadrant.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
2018-04-07 17:17:56 +02:00
case REORDER_BUFFER_CHANGE_TRUNCATE:
Fix memory leak in TRUNCATE decoding When decoding a TRUNCATE record, the relids array was being allocated in the main ReorderBuffer memory context, but not released with the change resulting in a memory leak. The array was also ignored when serializing/deserializing the change, assuming all the information is stored in the change itself. So when spilling the change to disk, we've only we have serialized only the pointer to the relids array. Thanks to never releasing the array, the pointer however remained valid even after loading the change back to memory, preventing an actual crash. This fixes both the memory leak and (de)serialization. The relids array is still allocated in the main ReorderBuffer memory context (none of the existing ones seems like a good match, and adding an extra context seems like an overkill). The allocation is wrapped in a new ReorderBuffer API functions, to keep the details within reorderbuffer.c, just like the other ReorderBufferGet methods do. Author: Tomas Vondra Discussion: https://www.postgresql.org/message-id/flat/66175a41-9342-2845-652f-1bd4c3ee50aa%402ndquadrant.com Backpatch: 11, where decoding of TRUNCATE was introduced
2018-09-03 02:10:24 +02:00
{
Oid *relids;
relids = ReorderBufferGetRelids(rb,
change->data.truncate.nrelids);
memcpy(relids, data, change->data.truncate.nrelids * sizeof(Oid));
change->data.truncate.relids = relids;
break;
}
Add support for INSERT ... ON CONFLICT DO NOTHING/UPDATE. The newly added ON CONFLICT clause allows to specify an alternative to raising a unique or exclusion constraint violation error when inserting. ON CONFLICT refers to constraints that can either be specified using a inference clause (by specifying the columns of a unique constraint) or by naming a unique or exclusion constraint. DO NOTHING avoids the constraint violation, without touching the pre-existing row. DO UPDATE SET ... [WHERE ...] updates the pre-existing tuple, and has access to both the tuple proposed for insertion and the existing tuple; the optional WHERE clause can be used to prevent an update from being executed. The UPDATE SET and WHERE clauses have access to the tuple proposed for insertion using the "magic" EXCLUDED alias, and to the pre-existing tuple using the table name or its alias. This feature is often referred to as upsert. This is implemented using a new infrastructure called "speculative insertion". It is an optimistic variant of regular insertion that first does a pre-check for existing tuples and then attempts an insert. If a violating tuple was inserted concurrently, the speculatively inserted tuple is deleted and a new attempt is made. If the pre-check finds a matching tuple the alternative DO NOTHING or DO UPDATE action is taken. If the insertion succeeds without detecting a conflict, the tuple is deemed inserted. To handle the possible ambiguity between the excluded alias and a table named excluded, and for convenience with long relation names, INSERT INTO now can alias its target table. Bumps catversion as stored rules change. Author: Peter Geoghegan, with significant contributions from Heikki Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes. Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs, Dean Rasheed, Stephen Frost and many others.
2015-05-08 05:31:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_CONFIRM:
Fix decoding of speculative aborts. During decoding for speculative inserts, we were relying for cleaning toast hash on confirmation records or next change records. But that could lead to multiple problems (a) memory leak if there is neither a confirmation record nor any other record after toast insertion for a speculative insert in the transaction, (b) error and assertion failures if the next operation is not an insert/update on the same table. The fix is to start queuing spec abort change and clean up toast hash and change record during its processing. Currently, we are queuing the spec aborts for both toast and main table even though we perform cleanup while processing the main table's spec abort record. Later, if we have a way to distinguish between the spec abort record of toast and the main table, we can avoid queuing the change for spec aborts of toast tables. Reported-by: Ashutosh Bapat Author: Dilip Kumar Reviewed-by: Amit Kapila Backpatch-through: 9.6, where it was introduced Discussion: https://postgr.es/m/CAExHW5sPKF-Oovx_qZe4p5oM6Dvof7_P+XgsNAViug15Fm99jA@mail.gmail.com
2021-06-15 04:58:36 +02:00
case REORDER_BUFFER_CHANGE_INTERNAL_SPEC_ABORT:
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
case REORDER_BUFFER_CHANGE_INTERNAL_COMMAND_ID:
case REORDER_BUFFER_CHANGE_INTERNAL_TUPLECID:
break;
}
dlist_push_tail(&txn->changes, &change->node);
txn->nentries_mem++;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
* Update memory accounting for the restored change. We need to do this
* although we don't check the memory limit when restoring the changes in
* this branch (we only do that when initially queueing the changes after
* decoding), because we will release the changes later, and that will
* update the accounting too (subtracting the size from the counters). And
* we don't want to underflow there.
*/
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
ReorderBufferChangeMemoryUpdate(rb, change, true,
ReorderBufferChangeSize(change));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Remove all on-disk stored for the passed in transaction.
*/
static void
ReorderBufferRestoreCleanup(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
XLogSegNo first;
XLogSegNo cur;
XLogSegNo last;
Assert(txn->first_lsn != InvalidXLogRecPtr);
Assert(txn->final_lsn != InvalidXLogRecPtr);
Make WAL segment size configurable at initdb time. For performance reasons a larger segment size than the default 16MB can be useful. A larger segment size has two main benefits: Firstly, in setups using archiving, it makes it easier to write scripts that can keep up with higher amounts of WAL, secondly, the WAL has to be written and synced to disk less frequently. But at the same time large segment size are disadvantageous for smaller databases. So far the segment size had to be configured at compile time, often making it unrealistic to choose one fitting to a particularly load. Therefore change it to a initdb time setting. This includes a breaking changes to the xlogreader.h API, which now requires the current segment size to be configured. For that and similar reasons a number of binaries had to be taught how to recognize the current segment size. Author: Beena Emerson, editorialized by Andres Freund Reviewed-By: Andres Freund, David Steele, Kuntal Ghosh, Michael Paquier, Peter Eisentraut, Robert Hass, Tushar Ahuja Discussion: https://postgr.es/m/CAOG9ApEAcQ--1ieKbhFzXSQPw_YLmepaa4hNdnY5+ZULpt81Mw@mail.gmail.com
2017-09-20 07:03:48 +02:00
XLByteToSeg(txn->first_lsn, first, wal_segment_size);
XLByteToSeg(txn->final_lsn, last, wal_segment_size);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* iterate over all possible filenames, and delete them */
for (cur = first; cur <= last; cur++)
{
char path[MAXPGPATH];
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
ReorderBufferSerializedPath(path, MyReplicationSlot, txn->xid, cur);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (unlink(path) != 0 && errno != ENOENT)
ereport(ERROR,
(errcode_for_file_access(),
Assorted message fixes and improvements
2014-09-05 07:20:33 +02:00
errmsg("could not remove file \"%s\": %m", path)));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
}
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
/*
* Remove any leftover serialized reorder buffers from a slot directory after a
* prior crash or decoding session exit.
*/
static void
ReorderBufferCleanupSerializedTXNs(const char *slotname)
{
DIR *spill_dir;
struct dirent *spill_de;
struct stat statbuf;
char path[MAXPGPATH * 2 + 12];
sprintf(path, "pg_replslot/%s", slotname);
/* we're only handling directories here, skip if it's not ours */
if (lstat(path, &statbuf) == 0 && !S_ISDIR(statbuf.st_mode))
return;
spill_dir = AllocateDir(path);
while ((spill_de = ReadDirExtended(spill_dir, path, INFO)) != NULL)
{
/* only look at names that can be ours */
if (strncmp(spill_de->d_name, "xid", 3) == 0)
{
snprintf(path, sizeof(path),
"pg_replslot/%s/%s", slotname,
spill_de->d_name);
if (unlink(path) != 0)
ereport(ERROR,
(errcode_for_file_access(),
Change extension of spilled ReorderBufferChange data to ".tmp". The previous extension, ".snap", was chosen for historical reasons and became confusing. Discussion: https://postgr.es/m/CAMp0ubd_P8vBGx8=MfDXQJZxHA5D_Zarw5cCkDxJ_63+pWRJ9w@mail.gmail.com
2018-08-25 18:19:21 +02:00
errmsg("could not remove file \"%s\" during removal of pg_replslot/%s/xid*: %m",
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
path, slotname)));
}
}
FreeDir(spill_dir);
}
/*
* Given a replication slot, transaction ID and segment number, fill in the
* corresponding spill file into 'path', which is a caller-owned buffer of size
* at least MAXPGPATH.
*/
static void
ReorderBufferSerializedPath(char *path, ReplicationSlot *slot, TransactionId xid,
XLogSegNo segno)
{
XLogRecPtr recptr;
Flip argument order in XLogSegNoOffsetToRecPtr Commit fc49e24fa69a added an input argument after the existing output argument. Flip those. Author: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Andres Freund <andres@anarazel.de> Discussion: https://postgr.es/m/20180708182345.imdgovmkffgtihhk@alvherre.pgsql
2018-07-09 20:28:21 +02:00
XLogSegNoOffsetToRecPtr(segno, 0, wal_segment_size, recptr);
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
Reconsider new file extension in commit 91f26d5f. Andres and Tom objected to the choice of the ".tmp" extension. Changing to Andres's suggestion of ".spill". Discussion: https://postgr.es/m/88092095-3348-49D8-8746-EB574B1D30EA%40anarazel.de
2018-08-26 07:45:59 +02:00
snprintf(path, MAXPGPATH, "pg_replslot/%s/xid-%u-lsn-%X-%X.spill",
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
NameStr(MyReplicationSlot->data.name),
Simplify printing of LSNs Add a macro LSN_FORMAT_ARGS for use in printf-style printing of LSNs. Convert all applicable code to use it. Reviewed-by: Ashutosh Bapat <ashutosh.bapat@enterprisedb.com> Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Discussion: https://www.postgresql.org/message-id/flat/CAExHW5ub5NaTELZ3hJUCE6amuvqAtsSxc7O+uK7y4t9Rrk23cw@mail.gmail.com
2021-02-23 10:14:38 +01:00
xid, LSN_FORMAT_ARGS(recptr));
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
}
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* Delete all data spilled to disk after we've restarted/crashed. It will be
* recreated when the respective slots are reused.
*/
void
StartupReorderBuffer(void)
{
DIR *logical_dir;
struct dirent *logical_de;
logical_dir = AllocateDir("pg_replslot");
while ((logical_de = ReadDir(logical_dir, "pg_replslot")) != NULL)
{
if (strcmp(logical_de->d_name, ".") == 0 ||
strcmp(logical_de->d_name, "..") == 0)
continue;
/* if it cannot be a slot, skip the directory */
if (!ReplicationSlotValidateName(logical_de->d_name, DEBUG2))
continue;
/*
* ok, has to be a surviving logical slot, iterate and delete
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
* everything starting with xid-*
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
Refrain from duplicating data in reorderbuffers If a walsender exits leaving data in reorderbuffers, the next walsender that tries to decode the same transaction would append its decoded data in the same spill files without truncating it first, which effectively duplicate the data. Avoid that by removing any leftover reorderbuffer spill files when a walsender starts. Backpatch to 9.4; this bug has been there from the very beginning of logical decoding. Author: Craig Ringer, revised by me Reviewed by: Álvaro Herrera, Petr Jelínek, Masahiko Sawada
2018-03-06 20:22:03 +01:00
ReorderBufferCleanupSerializedTXNs(logical_de->d_name);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
FreeDir(logical_dir);
}
/* ---------------------------------------
* toast reassembly support
* ---------------------------------------
*/
/*
* Initialize per tuple toast reconstruction support.
*/
static void
ReorderBufferToastInitHash(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
HASHCTL hash_ctl;
Assert(txn->toast_hash == NULL);
hash_ctl.keysize = sizeof(Oid);
hash_ctl.entrysize = sizeof(ReorderBufferToastEnt);
hash_ctl.hcxt = rb->context;
txn->toast_hash = hash_create("ReorderBufferToastHash", 5, &hash_ctl,
Improve hash_create's API for selecting simple-binary-key hash functions. Previously, if you wanted anything besides C-string hash keys, you had to specify a custom hashing function to hash_create(). Nearly all such callers were specifying tag_hash or oid_hash; which is tedious, and rather error-prone, since a caller could easily miss the opportunity to optimize by using hash_uint32 when appropriate. Replace this with a design whereby callers using simple binary-data keys just specify HASH_BLOBS and don't need to mess with specific support functions. hash_create() itself will take care of optimizing when the key size is four bytes. This nets out saving a few hundred bytes of code space, and offers a measurable performance improvement in tidbitmap.c (which was not exploiting the opportunity to use hash_uint32 for its 4-byte keys). There might be some wins elsewhere too, I didn't analyze closely. In future we could look into offering a similar optimized hashing function for 8-byte keys. Under this design that could be done in a centralized and machine-independent fashion, whereas getting it right for keys of platform-dependent sizes would've been notationally painful before. For the moment, the old way still works fine, so as not to break source code compatibility for loadable modules. Eventually we might want to remove tag_hash and friends from the exported API altogether, since there's no real need for them to be explicitly referenced from outside dynahash.c. Teodor Sigaev and Tom Lane
2014-12-18 19:36:29 +01:00
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Per toast-chunk handling for toast reconstruction
*
* Appends a toast chunk so we can reconstruct it when the tuple "owning" the
* toasted Datum comes along.
*/
static void
ReorderBufferToastAppendChunk(ReorderBuffer *rb, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change)
{
ReorderBufferToastEnt *ent;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferTupleBuf *newtup;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
bool found;
int32 chunksize;
bool isnull;
Pointer chunk;
TupleDesc desc = RelationGetDescr(relation);
Oid chunk_id;
Fix assorted confusion between Oid and int32. In passing, also make some debugging elog's in pgstat.c a bit more consistently worded. Back-patch as far as applicable (9.3 or 9.4; none of these mistakes are really old). Mark Dilger identified and patched the type violations; the message rewordings are mine.
2014-12-11 21:41:15 +01:00
int32 chunk_seq;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (txn->toast_hash == NULL)
ReorderBufferToastInitHash(rb, txn);
Assert(IsToastRelation(relation));
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
newtup = change->data.tp.newtuple;
chunk_id = DatumGetObjectId(fastgetattr(&newtup->tuple, 1, desc, &isnull));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(!isnull);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
chunk_seq = DatumGetInt32(fastgetattr(&newtup->tuple, 2, desc, &isnull));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(!isnull);
ent = (ReorderBufferToastEnt *)
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(txn->toast_hash, &chunk_id, HASH_ENTER, &found);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (!found)
{
Assert(ent->chunk_id == chunk_id);
ent->num_chunks = 0;
ent->last_chunk_seq = 0;
ent->size = 0;
ent->reconstructed = NULL;
dlist_init(&ent->chunks);
if (chunk_seq != 0)
elog(ERROR, "got sequence entry %d for toast chunk %u instead of seq 0",
chunk_seq, chunk_id);
}
else if (found && chunk_seq != ent->last_chunk_seq + 1)
elog(ERROR, "got sequence entry %d for toast chunk %u instead of seq %d",
chunk_seq, chunk_id, ent->last_chunk_seq + 1);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
chunk = DatumGetPointer(fastgetattr(&newtup->tuple, 3, desc, &isnull));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(!isnull);
/* calculate size so we can allocate the right size at once later */
if (!VARATT_IS_EXTENDED(chunk))
chunksize = VARSIZE(chunk) - VARHDRSZ;
else if (VARATT_IS_SHORT(chunk))
/* could happen due to heap_form_tuple doing its thing */
chunksize = VARSIZE_SHORT(chunk) - VARHDRSZ_SHORT;
else
elog(ERROR, "unexpected type of toast chunk");
ent->size += chunksize;
ent->last_chunk_seq = chunk_seq;
ent->num_chunks++;
dlist_push_tail(&ent->chunks, &change->node);
}
/*
* Rejigger change->newtuple to point to in-memory toast tuples instead to
* on-disk toast tuples that may not longer exist (think DROP TABLE or VACUUM).
*
* We cannot replace unchanged toast tuples though, so those will still point
* to on-disk toast data.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*
* While updating the existing change with detoasted tuple data, we need to
* update the memory accounting info, because the change size will differ.
* Otherwise the accounting may get out of sync, triggering serialization
* at unexpected times.
*
* We simply subtract size of the change before rejiggering the tuple, and
* then adding the new size. This makes it look like the change was removed
* and then added back, except it only tweaks the accounting info.
*
* In particular it can't trigger serialization, which would be pointless
* anyway as it happens during commit processing right before handing
* the change to the output plugin.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static void
ReorderBufferToastReplace(ReorderBuffer *rb, ReorderBufferTXN *txn,
Relation relation, ReorderBufferChange *change)
{
TupleDesc desc;
int natt;
Datum *attrs;
bool *isnull;
bool *free;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
HeapTuple tmphtup;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Relation toast_rel;
TupleDesc toast_desc;
MemoryContext oldcontext;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferTupleBuf *newtup;
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
Size old_size;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* no toast tuples changed */
if (txn->toast_hash == NULL)
return;
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/*
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
* We're going to modify the size of the change. So, to make sure the
* accounting is correct we record the current change size and then after
* re-computing the change we'll subtract the recorded size and then
* re-add the new change size at the end. We don't immediately subtract
* the old size because if there is any error before we add the new size,
* we will release the changes and that will update the accounting info
* (subtracting the size from the counters). And we don't want to
* underflow there.
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
*/
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
old_size = ReorderBufferChangeSize(change);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
oldcontext = MemoryContextSwitchTo(rb->context);
/* we should only have toast tuples in an INSERT or UPDATE */
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
Assert(change->data.tp.newtuple);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
desc = RelationGetDescr(relation);
toast_rel = RelationIdGetRelation(relation->rd_rel->reltoastrelid);
Fix RelationIdGetRelation calls that weren't bothering with error checks. Some of these are quite old, but that doesn't make them not bugs. We'd rather report a failure via elog than SIGSEGV. While at it, uniformly spell the error check as !RelationIsValid(rel) rather than a bare rel == NULL test. The machine code is the same but it seems better to be consistent. Coverity complained about this today, not sure why, because the mistake is in fact old.
2019-09-08 23:00:29 +02:00
if (!RelationIsValid(toast_rel))
Add parent table name in an error in reorderbuffer.c. This can help in troubleshooting the cause of a particular error that can occur during decoding. Author: Jeremy Schneider Reviewed-by: Amit Kapila Discussion: https://postgr.es/m/808ed65b-994c-915a-361c-577f088b837f@amazon.com
2021-09-22 04:12:52 +02:00
elog(ERROR, "could not open toast relation with OID %u (base relation \"%s\")",
relation->rd_rel->reltoastrelid, RelationGetRelationName(relation));
Fix RelationIdGetRelation calls that weren't bothering with error checks. Some of these are quite old, but that doesn't make them not bugs. We'd rather report a failure via elog than SIGSEGV. While at it, uniformly spell the error check as !RelationIsValid(rel) rather than a bare rel == NULL test. The machine code is the same but it seems better to be consistent. Coverity complained about this today, not sure why, because the mistake is in fact old.
2019-09-08 23:00:29 +02:00
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
toast_desc = RelationGetDescr(toast_rel);
/* should we allocate from stack instead? */
attrs = palloc0(sizeof(Datum) * desc->natts);
isnull = palloc0(sizeof(bool) * desc->natts);
free = palloc0(sizeof(bool) * desc->natts);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
newtup = change->data.tp.newtuple;
heap_deform_tuple(&newtup->tuple, desc, attrs, isnull);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
for (natt = 0; natt < desc->natts; natt++)
{
Change tupledesc->attrs[n] to TupleDescAttr(tupledesc, n). This is a mechanical change in preparation for a later commit that will change the layout of TupleDesc. Introducing a macro to abstract the details of where attributes are stored will allow us to change that in separate step and revise it in future. Author: Thomas Munro, editorialized by Andres Freund Reviewed-By: Andres Freund Discussion: https://postgr.es/m/CAEepm=0ZtQ-SpsgCyzzYpsXS6e=kZWqk3g5Ygn3MDV7A8dabUA@mail.gmail.com
2017-08-20 20:19:07 +02:00
Form_pg_attribute attr = TupleDescAttr(desc, natt);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ReorderBufferToastEnt *ent;
struct varlena *varlena;
/* va_rawsize is the size of the original datum -- including header */
struct varatt_external toast_pointer;
struct varatt_indirect redirect_pointer;
struct varlena *new_datum = NULL;
struct varlena *reconstructed;
dlist_iter it;
Size data_done = 0;
/* system columns aren't toasted */
if (attr->attnum < 0)
continue;
if (attr->attisdropped)
continue;
/* not a varlena datatype */
if (attr->attlen != -1)
continue;
/* no data */
if (isnull[natt])
continue;
/* ok, we know we have a toast datum */
varlena = (struct varlena *) DatumGetPointer(attrs[natt]);
/* no need to do anything if the tuple isn't external */
if (!VARATT_IS_EXTERNAL(varlena))
continue;
VARATT_EXTERNAL_GET_POINTER(toast_pointer, varlena);
/*
* Check whether the toast tuple changed, replace if so.
*/
ent = (ReorderBufferToastEnt *)
hash_search(txn->toast_hash,
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
&toast_pointer.va_valueid,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
HASH_FIND,
NULL);
if (ent == NULL)
continue;
new_datum =
(struct varlena *) palloc0(INDIRECT_POINTER_SIZE);
free[natt] = true;
reconstructed = palloc0(toast_pointer.va_rawsize);
ent->reconstructed = reconstructed;
/* stitch toast tuple back together from its parts */
dlist_foreach(it, &ent->chunks)
{
bool isnull;
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
ReorderBufferChange *cchange;
ReorderBufferTupleBuf *ctup;
Pointer chunk;
cchange = dlist_container(ReorderBufferChange, node, it.cur);
ctup = cchange->data.tp.newtuple;
chunk = DatumGetPointer(fastgetattr(&ctup->tuple, 3, toast_desc, &isnull));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Assert(!isnull);
Assert(!VARATT_IS_EXTERNAL(chunk));
Assert(!VARATT_IS_SHORT(chunk));
memcpy(VARDATA(reconstructed) + data_done,
VARDATA(chunk),
VARSIZE(chunk) - VARHDRSZ);
data_done += VARSIZE(chunk) - VARHDRSZ;
}
Allow configurable LZ4 TOAST compression. There is now a per-column COMPRESSION option which can be set to pglz (the default, and the only option in up until now) or lz4. Or, if you like, you can set the new default_toast_compression GUC to lz4, and then that will be the default for new table columns for which no value is specified. We don't have lz4 support in the PostgreSQL code, so to use lz4 compression, PostgreSQL must be built --with-lz4. In general, TOAST compression means compression of individual column values, not the whole tuple, and those values can either be compressed inline within the tuple or compressed and then stored externally in the TOAST table, so those properties also apply to this feature. Prior to this commit, a TOAST pointer has two unused bits as part of the va_extsize field, and a compessed datum has two unused bits as part of the va_rawsize field. These bits are unused because the length of a varlena is limited to 1GB; we now use them to indicate the compression type that was used. This means we only have bit space for 2 more built-in compresison types, but we could work around that problem, if necessary, by introducing a new vartag_external value for any further types we end up wanting to add. Hopefully, it won't be too important to offer a wide selection of algorithms here, since each one we add not only takes more coding but also adds a build dependency for every packager. Nevertheless, it seems worth doing at least this much, because LZ4 gets better compression than PGLZ with less CPU usage. It's possible for LZ4-compressed datums to leak into composite type values stored on disk, just as it is for PGLZ. It's also possible for LZ4-compressed attributes to be copied into a different table via SQL commands such as CREATE TABLE AS or INSERT .. SELECT. It would be expensive to force such values to be decompressed, so PostgreSQL has never done so. For the same reasons, we also don't force recompression of already-compressed values even if the target table prefers a different compression method than was used for the source data. These architectural decisions are perhaps arguable but revisiting them is well beyond the scope of what seemed possible to do as part of this project. However, it's relatively cheap to recompress as part of VACUUM FULL or CLUSTER, so this commit adjusts those commands to do so, if the configured compression method of the table happens not to match what was used for some column value stored therein. Dilip Kumar. The original patches on which this work was based were written by Ildus Kurbangaliev, and those were patches were based on even earlier work by Nikita Glukhov, but the design has since changed very substantially, since allow a potentially large number of compression methods that could be added and dropped on a running system proved too problematic given some of the architectural issues mentioned above; the choice of which specific compression method to add first is now different; and a lot of the code has been heavily refactored. More recently, Justin Przyby helped quite a bit with testing and reviewing and this version also includes some code contributions from him. Other design input and review from Tomas Vondra, Álvaro Herrera, Andres Freund, Oleg Bartunov, Alexander Korotkov, and me. Discussion: http://postgr.es/m/20170907194236.4cefce96%40wp.localdomain Discussion: http://postgr.es/m/CAFiTN-uUpX3ck%3DK0mLEk-G_kUQY%3DSNOTeqdaNRR9FMdQrHKebw%40mail.gmail.com
2021-03-19 20:10:38 +01:00
Assert(data_done == VARATT_EXTERNAL_GET_EXTSIZE(toast_pointer));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* make sure its marked as compressed or not */
if (VARATT_EXTERNAL_IS_COMPRESSED(toast_pointer))
SET_VARSIZE_COMPRESSED(reconstructed, data_done + VARHDRSZ);
else
SET_VARSIZE(reconstructed, data_done + VARHDRSZ);
memset(&redirect_pointer, 0, sizeof(redirect_pointer));
redirect_pointer.pointer = reconstructed;
SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
sizeof(redirect_pointer));
attrs[natt] = PointerGetDatum(new_datum);
}
/*
* Build tuple in separate memory & copy tuple back into the tuplebuf
* passed to the output plugin. We can't directly heap_fill_tuple() into
* the tuplebuf because attrs[] will point back into the current content.
*/
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
tmphtup = heap_form_tuple(desc, attrs, isnull);
Assert(newtup->tuple.t_len <= MaxHeapTupleSize);
logical decoding: Fix handling of large old tuples with replica identity full. When decoding the old version of an UPDATE or DELETE change, and if that tuple was bigger than MaxHeapTupleSize, we either Assert'ed out, or failed in more subtle ways in non-assert builds. Normally individual tuples aren't bigger than MaxHeapTupleSize, with big datums toasted. But that's not the case for the old version of a tuple for logical decoding; the replica identity is logged as one piece. With the default replica identity btree limits that to small tuples, but that's not the case for FULL. Change the tuple buffer infrastructure to separate allocate over-large tuples, instead of always going through the slab cache. This unfortunately requires changing the ReorderBufferTupleBuf definition, we need to store the allocated size someplace. To avoid requiring output plugins to recompile, don't store HeapTupleHeaderData directly after HeapTupleData, but point to it via t_data; that leaves rooms for the allocated size. As there's no reason for an output plugin to look at ReorderBufferTupleBuf->t_data.header, remove the field. It was just a minor convenience having it directly accessible. Reported-By: Adam Dratwiński Discussion: CAKg6ypLd7773AOX4DiOGRwQk1TVOQKhNwjYiVjJnpq8Wo+i62Q@mail.gmail.com
2016-03-06 03:02:20 +01:00
Assert(ReorderBufferTupleBufData(newtup) == newtup->tuple.t_data);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
memcpy(newtup->tuple.t_data, tmphtup->t_data, tmphtup->t_len);
newtup->tuple.t_len = tmphtup->t_len;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* free resources we won't further need, more persistent stuff will be
* free'd in ReorderBufferToastReset().
*/
RelationClose(toast_rel);
Remove unportable use of anonymous unions from reorderbuffer.h. In b89e151054a I had assumed it was ok to use anonymous unions as struct members, but while a longstanding extension in many compilers, it's only been standardized in C11. To fix, remove one of the anonymous unions which tried to hide some implementation specific enum values and give the other a name. The latter unfortunately requires changes in output plugins, but since the feature has only been added a few days ago... Andres Freund
2014-03-07 23:02:48 +01:00
pfree(tmphtup);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
for (natt = 0; natt < desc->natts; natt++)
{
if (free[natt])
pfree(DatumGetPointer(attrs[natt]));
}
pfree(attrs);
pfree(free);
pfree(isnull);
MemoryContextSwitchTo(oldcontext);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
/* subtract the old change size */
ReorderBufferChangeMemoryUpdate(rb, change, false, old_size);
Add logical_decoding_work_mem to limit ReorderBuffer memory usage. Instead of deciding to serialize a transaction merely based on the number of changes in that xact (toplevel or subxact), this makes the decisions based on amount of memory consumed by the changes. The memory limit is defined by a new logical_decoding_work_mem GUC, so for example we can do this SET logical_decoding_work_mem = '128kB' to reduce the memory usage of walsenders or set the higher value to reduce disk writes. The minimum value is 64kB. When adding a change to a transaction, we account for the size in two places. Firstly, in the ReorderBuffer, which is then used to decide if we reached the total memory limit. And secondly in the transaction the change belongs to, so that we can pick the largest transaction to evict (and serialize to disk). We still use max_changes_in_memory when loading changes serialized to disk. The trouble is we can't use the memory limit directly as there might be multiple subxact serialized, we need to read all of them but we don't know how many are there (and which subxact to read first). We do not serialize the ReorderBufferTXN entries, so if there is a transaction with many subxacts, most memory may be in this type of objects. Those records are not included in the memory accounting. We also do not account for INTERNAL_TUPLECID changes, which are kept in a separate list and not evicted from memory. Transactions with many CTID changes may consume significant amounts of memory, but we can't really do much about that. The current eviction algorithm is very simple - the transaction is picked merely by size, while it might be useful to also consider age (LSN) of the changes for example. With the new Generational memory allocator, evicting the oldest changes would make it more likely the memory gets actually pfreed. The logical_decoding_work_mem can be set in postgresql.conf, in which case it serves as the default for all publishers on that instance. Author: Tomas Vondra, with changes by Dilip Kumar and Amit Kapila Reviewed-by: Dilip Kumar and Amit Kapila Tested-By: Vignesh C Discussion: https://postgr.es/m/688b0b7f-2f6c-d827-c27b-216a8e3ea700@2ndquadrant.com
2019-11-16 13:19:33 +01:00
/* now add the change back, with the correct size */
Fix reorder buffer memory accounting for toast changes. While processing toast changes in logical decoding, we rejigger the tuple change to point to in-memory toast tuples instead to on-disk toast tuples. And, to make sure the memory accounting is correct, we were subtracting the old change size and then after re-computing the new tuple, re-adding its size at the end. Now, if there is any error before we add the new size, we will release the changes and that will update the accounting info (subtracting the size from the counters). And we were underflowing there which leads to an assertion failure in assert enabled builds and wrong memory accounting in reorder buffer otherwise. Author: Bertrand Drouvot Reviewed-by: Amit Kapila Backpatch-through: 13, where memory accounting was introduced Discussion: https://postgr.es/m/92b0ee65-b8bd-e42d-c082-4f3f4bf12d34@amazon.com
2021-09-13 06:54:00 +02:00
ReorderBufferChangeMemoryUpdate(rb, change, true,
ReorderBufferChangeSize(change));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
* Free all resources allocated for toast reconstruction.
*/
static void
ReorderBufferToastReset(ReorderBuffer *rb, ReorderBufferTXN *txn)
{
HASH_SEQ_STATUS hstat;
ReorderBufferToastEnt *ent;
if (txn->toast_hash == NULL)
return;
/* sequentially walk over the hash and free everything */
hash_seq_init(&hstat, txn->toast_hash);
while ((ent = (ReorderBufferToastEnt *) hash_seq_search(&hstat)) != NULL)
{
dlist_mutable_iter it;
if (ent->reconstructed != NULL)
pfree(ent->reconstructed);
dlist_foreach_modify(it, &ent->chunks)
{
ReorderBufferChange *change =
dlist_container(ReorderBufferChange, node, it.cur);
dlist_delete(&change->node);
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
ReorderBufferReturnChange(rb, change, true);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
}
hash_destroy(txn->toast_hash);
txn->toast_hash = NULL;
}
/* ---------------------------------------
* Visibility support for logical decoding
*
*
* Lookup actual cmin/cmax values when using decoding snapshot. We can't
* always rely on stored cmin/cmax values because of two scenarios:
*
* * A tuple got changed multiple times during a single transaction and thus
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* has got a combo CID. Combo CIDs are only valid for the duration of a
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* single transaction.
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* * A tuple with a cmin but no cmax (and thus no combo CID) got
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* deleted/updated in another transaction than the one which created it
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* which we are looking at right now. As only one of cmin, cmax or combo CID
Fix some typos introduced by the logical decoding patch. Erik Rijkers
2014-03-05 19:00:22 +01:00
* is actually stored in the heap we don't have access to the value we
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* need anymore.
*
* To resolve those problems we have a per-transaction hash of (cmin,
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
* cmax) tuples keyed by (relfilelocator, ctid) which contains the actual
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* (cmin, cmax) values. That also takes care of combo CIDs by simply
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
* not caring about them at all. As we have the real cmin/cmax values
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* combo CIDs aren't interesting.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*
* As we only care about catalog tuples here the overhead of this
* hashtable should be acceptable.
*
* Heap rewrites complicate this a bit, check rewriteheap.c for
* details.
* -------------------------------------------------------------------------
*/
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
/* struct for sorting mapping files by LSN efficiently */
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
typedef struct RewriteMappingFile
{
XLogRecPtr lsn;
char fname[MAXPGPATH];
} RewriteMappingFile;
Fix most -Wundef warnings In some cases #if was used instead of #ifdef in an inconsistent style. Cleaning this up also helps when analyzing cases like 38d8dce61fff09daae0edb6bcdd42b0c7f10ebcd where this makes a difference. There are no behavior changes here, but the change in pg_bswap.h would prevent possible accidental misuse by third-party code. Discussion: https://www.postgresql.org/message-id/flat/3b615ca5-c595-3f1d-fdf7-a429e564f614%402ndquadrant.com
2019-10-19 18:21:58 +02:00
#ifdef NOT_USED
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
static void
DisplayMapping(HTAB *tuplecid_data)
{
HASH_SEQ_STATUS hstat;
ReorderBufferTupleCidEnt *ent;
hash_seq_init(&hstat, tuplecid_data);
while ((ent = (ReorderBufferTupleCidEnt *) hash_seq_search(&hstat)) != NULL)
{
Revert 56-bit relfilenode change and follow-up commits. There are still some alignment-related failures in the buildfarm, which might or might not be able to be fixed quickly, but I've also just realized that it increased the size of many WAL records by 4 bytes because a block reference contains a RelFileLocator. The effect of that hasn't been studied or discussed, so revert for now.
2022-09-28 15:45:27 +02:00
elog(DEBUG3, "mapping: node: %u/%u/%u tid: %u/%u cmin: %u, cmax: %u",
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
ent->key.rlocator.dbOid,
ent->key.rlocator.spcOid,
ent->key.rlocator.relNumber,
Remove direct uses of ItemPointer.{ip_blkid,ip_posid} There are no functional changes here; this simply encapsulates knowledge of the ItemPointerData struct so that a future patch can change things without more breakage. All direct users of ip_blkid and ip_posid are changed to use existing macros ItemPointerGetBlockNumber and ItemPointerGetOffsetNumber respectively. For callers where that's inappropriate (because they Assert that the itempointer is is valid-looking), add ItemPointerGetBlockNumberNoCheck and ItemPointerGetOffsetNumberNoCheck, which lack the assertion but are otherwise identical. Author: Pavan Deolasee Discussion: https://postgr.es/m/CABOikdNnFon4cJiL=h1mZH3bgUeU+sWHuU4Yr8AB=j3A2p1GiA@mail.gmail.com
2017-03-28 17:52:55 +02:00
ItemPointerGetBlockNumber(&ent->key.tid),
ItemPointerGetOffsetNumber(&ent->key.tid),
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ent->cmin,
ent->cmax
);
}
}
#endif
/*
* Apply a single mapping file to tuplecid_data.
*
* The mapping file has to have been verified to be a) committed b) for our
* transaction c) applied in LSN order.
*/
static void
ApplyLogicalMappingFile(HTAB *tuplecid_data, Oid relid, const char *fname)
{
char path[MAXPGPATH];
int fd;
int readBytes;
LogicalRewriteMappingData map;
Rename logical decoding's pg_llog directory to pg_logical. The old name wasn't very descriptive as of actual contents of the directory, which are historical snapshots in the snapshots/ subdirectory and mappingdata for rewritten tuples in mappings/. There's been a fair amount of discussion what would be a good name. I'm settling for pg_logical because it's likely that further data around logical decoding and replication will need saving in the future. Also add the missing entry for the directory into storage.sgml's list of PGDATA contents. Bumps catversion as the data directories won't be compatible.
2014-07-02 21:07:47 +02:00
sprintf(path, "pg_logical/mappings/%s", fname);
Refactor new file permission handling The file handling functions from fd.c were called with a diverse mix of notations for the file permissions when they were opening new files. Almost all files created by the server should have the same permissions set. So change the API so that e.g. OpenTransientFile() automatically uses the standard permissions set, and OpenTransientFilePerm() is a new function that takes an explicit permissions set for the few cases where it is needed. This also saves an unnecessary argument for call sites that are just opening an existing file. While we're reviewing these APIs, get rid of the FileName typedef and use the standard const char * for the file name and mode_t for the file mode. This makes these functions match other file handling functions and removes an unnecessary layer of mysteriousness. We can also get rid of a few casts that way. Author: David Steele <david@pgmasters.net>
2017-09-23 15:49:22 +02:00
fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (fd < 0)
ereport(ERROR,
reorderbuffer: preserve errno while reporting error Clobbering errno during cleanup after an error is an oft-repeated, easy to make mistake. Deal with it here as everywhere else, by saving it aside and restoring after cleanup, before ereport'ing. In passing, add a missing errcode declaration in another ereport() call in the same file, which I noticed while skimming the file looking for similar problems. Backpatch to 9.4, where this code was introduced.
2016-08-19 19:38:55 +02:00
(errcode_for_file_access(),
errmsg("could not open file \"%s\": %m", path)));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
while (true)
{
ReorderBufferTupleCidKey key;
ReorderBufferTupleCidEnt *ent;
ReorderBufferTupleCidEnt *new_ent;
bool found;
/* be careful about padding */
memset(&key, 0, sizeof(ReorderBufferTupleCidKey));
/* read all mappings till the end of the file */
Create and use wait events for read, write, and fsync operations. Previous commits, notably 53be0b1add7064ca5db3cd884302dfc3268d884e and 6f3bd98ebfc008cbd676da777bb0b2376c4c4bfa, made it possible to see from pg_stat_activity when a backend was stuck waiting for another backend, but it's also fairly common for a backend to be stuck waiting for an I/O. Add wait events for those operations, too. Rushabh Lathia, with further hacking by me. Reviewed and tested by Michael Paquier, Amit Kapila, Rajkumar Raghuwanshi, and Rahila Syed. Discussion: http://postgr.es/m/CAGPqQf0LsYHXREPAZqYGVkDqHSyjf=KsD=k0GTVPAuzyThh-VQ@mail.gmail.com
2017-03-18 12:43:01 +01:00
pgstat_report_wait_start(WAIT_EVENT_REORDER_LOGICAL_MAPPING_READ);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
readBytes = read(fd, &map, sizeof(LogicalRewriteMappingData));
Create and use wait events for read, write, and fsync operations. Previous commits, notably 53be0b1add7064ca5db3cd884302dfc3268d884e and 6f3bd98ebfc008cbd676da777bb0b2376c4c4bfa, made it possible to see from pg_stat_activity when a backend was stuck waiting for another backend, but it's also fairly common for a backend to be stuck waiting for an I/O. Add wait events for those operations, too. Rushabh Lathia, with further hacking by me. Reviewed and tested by Michael Paquier, Amit Kapila, Rajkumar Raghuwanshi, and Rahila Syed. Discussion: http://postgr.es/m/CAGPqQf0LsYHXREPAZqYGVkDqHSyjf=KsD=k0GTVPAuzyThh-VQ@mail.gmail.com
2017-03-18 12:43:01 +01:00
pgstat_report_wait_end();
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (readBytes < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not read file \"%s\": %m",
path)));
else if (readBytes == 0) /* EOF */
break;
else if (readBytes != sizeof(LogicalRewriteMappingData))
ereport(ERROR,
(errcode_for_file_access(),
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
errmsg("could not read from file \"%s\": read %d instead of %d bytes",
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
path, readBytes,
(int32) sizeof(LogicalRewriteMappingData))));
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
key.rlocator = map.old_locator;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ItemPointerCopy(&map.old_tid,
&key.tid);
ent = (ReorderBufferTupleCidEnt *)
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(tuplecid_data, &key, HASH_FIND, NULL);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* no existing mapping, no need to update */
if (!ent)
continue;
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
key.rlocator = map.new_locator;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
ItemPointerCopy(&map.new_tid,
&key.tid);
new_ent = (ReorderBufferTupleCidEnt *)
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(tuplecid_data, &key, HASH_ENTER, &found);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (found)
{
/*
* Make sure the existing mapping makes sense. We sometime update
* old records that did not yet have a cmax (e.g. pg_class' own
* entry while rewriting it) during rewrites, so allow that.
*/
Assert(ent->cmin == InvalidCommandId || ent->cmin == new_ent->cmin);
Assert(ent->cmax == InvalidCommandId || ent->cmax == new_ent->cmax);
}
else
{
/* update mapping */
new_ent->cmin = ent->cmin;
new_ent->cmax = ent->cmax;
new_ent->combocid = ent->combocid;
}
}
Close the file descriptor in ApplyLogicalMappingFile The function was forgetting to close the file descriptor, resulting in failures like this: ERROR: 53000: exceeded maxAllocatedDescs (492) while trying to open file "pg_logical/mappings/map-4000-4eb-1_60DE1E08-5376b5-537c6b" LOCATION: OpenTransientFile, fd.c:2161 Simply close the file at the end, and backpatch to 9.4 (where logical decoding was introduced). While at it, fix a nearby typo. Discussion: https://www.postgresql.org/message-id/flat/738a590a-2ce5-9394-2bef-7b1caad89b37%402ndquadrant.com
2018-08-16 16:49:10 +02:00
Use consistent style for checking return from system calls Use if (something() != 0) error ... instead of just if (something) error ... The latter is not incorrect, but it's a bit confusing and not the common style. Discussion: https://www.postgresql.org/message-id/flat/5de61b6b-8be9-7771-0048-860328efe027%402ndquadrant.com
2019-07-06 23:18:46 +02:00
if (CloseTransientFile(fd) != 0)
Tighten use of OpenTransientFile and CloseTransientFile This fixes two sets of issues related to the use of transient files in the backend: 1) OpenTransientFile() has been used in some code paths with read-write flags while read-only is sufficient, so switch those calls to be read-only where necessary. These have been reported by Joe Conway. 2) When opening transient files, it is up to the caller to close the file descriptors opened. In error code paths, CloseTransientFile() gets called to clean up things before issuing an error. However in normal exit paths, a lot of callers of CloseTransientFile() never actually reported errors, which could leave a file descriptor open without knowing about it. This is an issue I complained about a couple of times, but never had the courage to write and submit a patch, so here we go. Note that one frontend code path is impacted by this commit so as an error is issued when fetching control file data, making backend and frontend to be treated consistently. Reported-by: Joe Conway, Michael Paquier Author: Michael Paquier Reviewed-by: Álvaro Herrera, Georgios Kokolatos, Joe Conway Discussion: https://postgr.es/m/20190301023338.GD1348@paquier.xyz Discussion: https://postgr.es/m/c49b69ec-e2f7-ff33-4f17-0eaa4f2cef27@joeconway.com
2019-03-09 00:50:55 +01:00
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not close file \"%s\": %m", path)));
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
}
/*
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
* Check whether the TransactionId 'xid' is in the pre-sorted array 'xip'.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static bool
TransactionIdInArray(TransactionId xid, TransactionId *xip, Size num)
{
return bsearch(&xid, xip, num,
sizeof(TransactionId), xidComparator) != NULL;
}
/*
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
* list_sort() comparator for sorting RewriteMappingFiles in LSN order.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
static int
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
file_sort_by_lsn(const ListCell *a_p, const ListCell *b_p)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
RewriteMappingFile *a = (RewriteMappingFile *) lfirst(a_p);
RewriteMappingFile *b = (RewriteMappingFile *) lfirst(b_p);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
if (a->lsn < b->lsn)
return -1;
else if (a->lsn > b->lsn)
return 1;
return 0;
}
/*
* Apply any existing logical remapping files if there are any targeted at our
* transaction for relid.
*/
static void
UpdateLogicalMappings(HTAB *tuplecid_data, Oid relid, Snapshot snapshot)
{
DIR *mapping_dir;
struct dirent *mapping_de;
List *files = NIL;
ListCell *file;
Oid dboid = IsSharedRelation(relid) ? InvalidOid : MyDatabaseId;
Rename logical decoding's pg_llog directory to pg_logical. The old name wasn't very descriptive as of actual contents of the directory, which are historical snapshots in the snapshots/ subdirectory and mappingdata for rewritten tuples in mappings/. There's been a fair amount of discussion what would be a good name. I'm settling for pg_logical because it's likely that further data around logical decoding and replication will need saving in the future. Also add the missing entry for the directory into storage.sgml's list of PGDATA contents. Bumps catversion as the data directories won't be compatible.
2014-07-02 21:07:47 +02:00
mapping_dir = AllocateDir("pg_logical/mappings");
while ((mapping_de = ReadDir(mapping_dir, "pg_logical/mappings")) != NULL)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Oid f_dboid;
Oid f_relid;
TransactionId f_mapped_xid;
TransactionId f_create_xid;
XLogRecPtr f_lsn;
uint32 f_hi,
f_lo;
RewriteMappingFile *f;
if (strcmp(mapping_de->d_name, ".") == 0 ||
strcmp(mapping_de->d_name, "..") == 0)
continue;
/* Ignore files that aren't ours */
if (strncmp(mapping_de->d_name, "map-", 4) != 0)
continue;
if (sscanf(mapping_de->d_name, LOGICAL_REWRITE_FORMAT,
&f_dboid, &f_relid, &f_hi, &f_lo,
&f_mapped_xid, &f_create_xid) != 6)
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
elog(ERROR, "could not parse filename \"%s\"", mapping_de->d_name);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
f_lsn = ((uint64) f_hi) << 32 | f_lo;
/* mapping for another database */
if (f_dboid != dboid)
continue;
/* mapping for another relation */
if (f_relid != relid)
continue;
/* did the creating transaction abort? */
if (!TransactionIdDidCommit(f_create_xid))
continue;
/* not for our transaction */
if (!TransactionIdInArray(f_mapped_xid, snapshot->subxip, snapshot->subxcnt))
continue;
/* ok, relevant, queue for apply */
f = palloc(sizeof(RewriteMappingFile));
f->lsn = f_lsn;
strcpy(f->fname, mapping_de->d_name);
files = lappend(files, f);
}
FreeDir(mapping_dir);
/* sort files so we apply them in LSN order */
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
list_sort(files, file_sort_by_lsn);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
foreach(file, files)
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
{
Redesign the API for list sorting (list_qsort becomes list_sort). In the wake of commit 1cff1b95a, the obvious way to sort a List is to apply qsort() directly to the array of ListCells. list_qsort was building an intermediate array of pointers-to-ListCells, which we no longer need, but getting rid of it forces an API change: the comparator functions need to do one less level of indirection. Since we're having to touch the callers anyway, let's do two additional changes: sort the given list in-place rather than making a copy (as none of the existing callers have any use for the copying behavior), and rename list_qsort to list_sort. It was argued that the old name exposes more about the implementation than it should, which I find pretty questionable, but a better reason to rename it is to be sure we get the attention of any external callers about the need to fix their comparator functions. While we're at it, change four existing callers of qsort() to use list_sort instead; previously, they all had local reinventions of list_qsort, ie build-an-array-from-a-List-and-qsort-it. (There are some other places where changing to list_sort perhaps would be worthwhile, but they're less obviously wins.) Discussion: https://postgr.es/m/29361.1563220190@sss.pgh.pa.us
2019-07-16 17:51:44 +02:00
RewriteMappingFile *f = (RewriteMappingFile *) lfirst(file);
pgindent run for 9.4 This includes removing tabs after periods in C comments, which was applied to back branches, so this change should not effect backpatching.
2014-05-06 18:12:18 +02:00
Improve error messages in reorderbuffer.c. Be more clear about failure cases in relfilenode->relation lookup, and fix some other places that were inconsistent or not per our message style guidelines. Andres Freund and Tom Lane
2014-05-01 00:16:53 +02:00
elog(DEBUG1, "applying mapping: \"%s\" in %u", f->fname,
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
snapshot->subxip[0]);
ApplyLogicalMappingFile(tuplecid_data, relid, f->fname);
pfree(f);
}
}
/*
* Lookup cmin/cmax of a tuple, during logical decoding where we can't rely on
Sanitize the term "combo CID" in code comments Combo CIDs were referred in the code comments using different terms across various places of the code, so unify a bit the term used with what is currently in use in some of the READMEs. Author: "Hou, Zhijie" Discussion: https://postgr.es/m/1d42865c91404f46af4562532fdbea31@G08CNEXMBPEKD05.g08.fujitsu.local
2021-03-25 08:08:03 +01:00
* combo CIDs.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
*/
bool
ResolveCminCmaxDuringDecoding(HTAB *tuplecid_data,
Snapshot snapshot,
HeapTuple htup, Buffer buffer,
CommandId *cmin, CommandId *cmax)
{
ReorderBufferTupleCidKey key;
ReorderBufferTupleCidEnt *ent;
ForkNumber forkno;
BlockNumber blockno;
bool updated_mapping = false;
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
/*
* Return unresolved if tuplecid_data is not valid. That's because when
* streaming in-progress transactions we may run into tuples with the CID
* before actually decoding them. Think e.g. about INSERT followed by
* TRUNCATE, where the TRUNCATE may not be decoded yet when applying the
* INSERT. So in such cases, we assume the CID is from the future
* command.
*/
if (tuplecid_data == NULL)
return false;
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* be careful about padding */
memset(&key, 0, sizeof(key));
Assert(!BufferIsLocal(buffer));
/*
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
* get relfilelocator from the buffer, no convenient way to access it
* other than that.
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +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
BufferGetTag(buffer, &key.rlocator, &forkno, &blockno);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/* tuples can only be in the main fork */
Assert(forkno == MAIN_FORKNUM);
Assert(blockno == ItemPointerGetBlockNumber(&htup->t_self));
ItemPointerCopy(&htup->t_self,
&key.tid);
restart:
ent = (ReorderBufferTupleCidEnt *)
Remove useless casts to (void *) in hash_search() calls Some of these appear to be leftovers from when hash_search() took a char * argument (changed in 5999e78fc45dcb91784b64b6e9ae43f4e4f68ca2). Since after this there is some more horizontal space available, do some light reformatting where suitable. Reviewed-by: Corey Huinker <corey.huinker@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/fd9adf5d-b1aa-e82f-e4c7-263c30145807%40enterprisedb.com
2023-02-06 09:05:20 +01:00
hash_search(tuplecid_data, &key, HASH_FIND, NULL);
Introduce logical decoding. This feature, building on previous commits, allows the write-ahead log stream to be decoded into a series of logical changes; that is, inserts, updates, and deletes and the transactions which contain them. It is capable of handling decoding even across changes to the schema of the effected tables. The output format is controlled by a so-called "output plugin"; an example is included. To make use of this in a real replication system, the output plugin will need to be modified to produce output in the format appropriate to that system, and to perform filtering. Currently, information can be extracted from the logical decoding system only via SQL; future commits will add the ability to stream changes via walsender. Andres Freund, with review and other contributions from many other people, including Álvaro Herrera, Abhijit Menon-Sen, Peter Gheogegan, Kevin Grittner, Robert Haas, Heikki Linnakangas, Fujii Masao, Abhijit Menon-Sen, Michael Paquier, Simon Riggs, Craig Ringer, and Steve Singer.
2014-03-03 22:32:18 +01:00
/*
* failed to find a mapping, check whether the table was rewritten and
* apply mapping if so, but only do that once - there can be no new
* mappings while we are in here since we have to hold a lock on the
* relation.
*/
if (ent == NULL && !updated_mapping)
{
UpdateLogicalMappings(tuplecid_data, htup->t_tableOid, snapshot);
/* now check but don't update for a mapping again */
updated_mapping = true;
goto restart;
}
else if (ent == NULL)
return false;
if (cmin)
*cmin = ent->cmin;
if (cmax)
*cmax = ent->cmax;
return true;
}