rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
postgresql/contrib/amcheck/verify_nbtree.c

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

3263 lines
121 KiB
C
Raw Normal View History

Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*-------------------------------------------------------------------------
*
* verify_nbtree.c
* Verifies the integrity of nbtree indexes based on invariants.
*
* For B-Tree indexes, verification includes checking that each page in the
* target index has items in logical order as reported by an insertion scankey
* (the insertion scankey sort-wise NULL semantics are needed for
* verification).
*
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* When index-to-heap verification is requested, a Bloom filter is used to
* fingerprint all tuples in the target index, as the index is traversed to
* verify its structure. A heap scan later uses Bloom filter probes to verify
* that every visible heap tuple has a matching index tuple.
*
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Update copyright for 2023 Backpatch-through: 11
2023-01-02 21:00:37 +01:00
* Copyright (c) 2017-2023, PostgreSQL Global Development Group
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* IDENTIFICATION
* contrib/amcheck/verify_nbtree.c
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
#include "access/htup_details.h"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
#include "access/nbtree.h"
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
#include "access/table.h"
tableam: Add and use scan APIs. Too allow table accesses to be not directly dependent on heap, several new abstractions are needed. Specifically: 1) Heap scans need to be generalized into table scans. Do this by introducing TableScanDesc, which will be the "base class" for individual AMs. This contains the AM independent fields from HeapScanDesc. The previous heap_{beginscan,rescan,endscan} et al. have been replaced with a table_ version. There's no direct replacement for heap_getnext(), as that returned a HeapTuple, which is undesirable for a other AMs. Instead there's table_scan_getnextslot(). But note that heap_getnext() lives on, it's still used widely to access catalog tables. This is achieved by new scan_begin, scan_end, scan_rescan, scan_getnextslot callbacks. 2) The portion of parallel scans that's shared between backends need to be able to do so without the user doing per-AM work. To achieve that new parallelscan_{estimate, initialize, reinitialize} callbacks are introduced, which operate on a new ParallelTableScanDesc, which again can be subclassed by AMs. As it is likely that several AMs are going to be block oriented, block oriented callbacks that can be shared between such AMs are provided and used by heap. table_block_parallelscan_{estimate, intiialize, reinitialize} as callbacks, and table_block_parallelscan_{nextpage, init} for use in AMs. These operate on a ParallelBlockTableScanDesc. 3) Index scans need to be able to access tables to return a tuple, and there needs to be state across individual accesses to the heap to store state like buffers. That's now handled by introducing a sort-of-scan IndexFetchTable, which again is intended to be subclassed by individual AMs (for heap IndexFetchHeap). The relevant callbacks for an AM are index_fetch_{end, begin, reset} to create the necessary state, and index_fetch_tuple to retrieve an indexed tuple. Note that index_fetch_tuple implementations need to be smarter than just blindly fetching the tuples for AMs that have optimizations similar to heap's HOT - the currently alive tuple in the update chain needs to be fetched if appropriate. Similar to table_scan_getnextslot(), it's undesirable to continue to return HeapTuples. Thus index_fetch_heap (might want to rename that later) now accepts a slot as an argument. Core code doesn't have a lot of call sites performing index scans without going through the systable_* API (in contrast to loads of heap_getnext calls and working directly with HeapTuples). Index scans now store the result of a search in IndexScanDesc->xs_heaptid, rather than xs_ctup->t_self. As the target is not generally a HeapTuple anymore that seems cleaner. To be able to sensible adapt code to use the above, two further callbacks have been introduced: a) slot_callbacks returns a TupleTableSlotOps* suitable for creating slots capable of holding a tuple of the AMs type. table_slot_callbacks() and table_slot_create() are based upon that, but have additional logic to deal with views, foreign tables, etc. While this change could have been done separately, nearly all the call sites that needed to be adapted for the rest of this commit also would have been needed to be adapted for table_slot_callbacks(), making separation not worthwhile. b) tuple_satisfies_snapshot checks whether the tuple in a slot is currently visible according to a snapshot. That's required as a few places now don't have a buffer + HeapTuple around, but a slot (which in heap's case internally has that information). Additionally a few infrastructure changes were needed: I) SysScanDesc, as used by systable_{beginscan, getnext} et al. now internally uses a slot to keep track of tuples. While systable_getnext() still returns HeapTuples, and will so for the foreseeable future, the index API (see 1) above) now only deals with slots. The remainder, and largest part, of this commit is then adjusting all scans in postgres to use the new APIs. Author: Andres Freund, Haribabu Kommi, Alvaro Herrera Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de https://postgr.es/m/20160812231527.GA690404@alvherre.pgsql
2019-03-11 20:46:41 +01:00
#include "access/tableam.h"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
#include "access/transam.h"
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
#include "access/xact.h"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
#include "catalog/index.h"
#include "catalog/pg_am.h"
#include "commands/tablecmds.h"
Replace random(), pg_erand48(), etc with a better PRNG API and algorithm. Standardize on xoroshiro128** as our basic PRNG algorithm, eliminating a bunch of platform dependencies as well as fundamentally-obsolete PRNG code. In addition, this API replacement will ease replacing the algorithm again in future, should that become necessary. xoroshiro128** is a few percent slower than the drand48 family, but it can produce full-width 64-bit random values not only 48-bit, and it should be much more trustworthy. It's likely to be noticeably faster than the platform's random(), depending on which platform you are thinking about; and we can have non-global state vectors easily, unlike with random(). It is not cryptographically strong, but neither are the functions it replaces. Fabien Coelho, reviewed by Dean Rasheed, Aleksander Alekseev, and myself Discussion: https://postgr.es/m/alpine.DEB.2.22.394.2105241211230.165418@pseudo
2021-11-29 03:32:36 +01:00
#include "common/pg_prng.h"
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
#include "lib/bloomfilter.h"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
#include "miscadmin.h"
#include "storage/lmgr.h"
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
#include "storage/smgr.h"
Split up guc.c for better build speed and ease of maintenance. guc.c has grown to be one of our largest .c files, making it a bottleneck for compilation. It's also acquired a bunch of knowledge that'd be better kept elsewhere, because of our not very good habit of putting variable-specific check hooks here. Hence, split it up along these lines: * guc.c itself retains just the core GUC housekeeping mechanisms. * New file guc_funcs.c contains the SET/SHOW interfaces and some SQL-accessible functions for GUC manipulation. * New file guc_tables.c contains the data arrays that define the built-in GUC variables, along with some already-exported constant tables. * GUC check/assign/show hook functions are moved to the variable's home module, whenever that's clearly identifiable. A few hard- to-classify hooks ended up in commands/variable.c, which was already a home for miscellaneous GUC hook functions. To avoid cluttering a lot more header files with #include "guc.h", I also invented a new header file utils/guc_hooks.h and put all the GUC hook functions' declarations there, regardless of their originating module. That allowed removal of #include "guc.h" from some existing headers. The fallout from that (hopefully all caught here) demonstrates clearly why such inclusions are best minimized: there are a lot of files that, for example, were getting array.h at two or more levels of remove, despite not having any connection at all to GUCs in themselves. There is some very minor code beautification here, such as renaming a couple of inconsistently-named hook functions and improving some comments. But mostly this just moves code from point A to point B and deals with the ensuing needs for #include adjustments and exporting a few functions that previously weren't exported. Patch by me, per a suggestion from Andres Freund; thanks also to Michael Paquier for the idea to invent guc_funcs.c. Discussion: https://postgr.es/m/587607.1662836699@sss.pgh.pa.us
2022-09-13 17:05:07 +02:00
#include "utils/guc.h"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
#include "utils/memutils.h"
#include "utils/snapmgr.h"
PG_MODULE_MAGIC;
/*
* A B-Tree cannot possibly have this many levels, since there must be one
* block per level, which is bound by the range of BlockNumber:
*/
#define InvalidBtreeLevel ((uint32) InvalidBlockNumber)
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
#define BTreeTupleGetNKeyAtts(itup, rel) \
Min(IndexRelationGetNumberOfKeyAttributes(rel), BTreeTupleGetNAtts(itup, rel))
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* State associated with verifying a B-Tree index
*
* target is the point of reference for a verification operation.
*
* Other B-Tree pages may be allocated, but those are always auxiliary (e.g.,
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* they are current target's child pages). Conceptually, problems are only
* ever found in the current target page (or for a particular heap tuple during
* heapallindexed verification). Each page found by verification's left/right,
* top/bottom scan becomes the target exactly once.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
typedef struct BtreeCheckState
{
/*
* Unchanging state, established at start of verification:
*/
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* B-Tree Index Relation and associated heap relation */
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Relation rel;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Relation heaprel;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/* rel is heapkeyspace index? */
bool heapkeyspace;
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
/* ShareLock held on heap/index, rather than AccessShareLock? */
bool readonly;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* Also verifying heap has no unindexed tuples? */
bool heapallindexed;
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
/* Also making sure non-pivot tuples can be found by new search? */
bool rootdescend;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Per-page context */
MemoryContext targetcontext;
/* Buffer access strategy */
BufferAccessStrategy checkstrategy;
/*
* Mutable state, for verification of particular page:
*/
/* Current target page */
Page target;
/* Target block number */
BlockNumber targetblock;
/* Target page's LSN */
XLogRecPtr targetlsn;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/*
* Low key: high key of left sibling of target page. Used only for child
* verification. So, 'lowkey' is kept only when 'readonly' is set.
*/
IndexTuple lowkey;
/*
* The rightlink and incomplete split flag of block one level down to the
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
* target page, which was visited last time via downlink from target page.
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* We use it to check for missing downlinks.
*/
BlockNumber prevrightlink;
bool previncompletesplit;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/*
* Mutable state, for optional heapallindexed verification:
*/
/* Bloom filter fingerprints B-Tree index */
bloom_filter *filter;
/* Debug counter */
int64 heaptuplespresent;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
} BtreeCheckState;
/*
* Starting point for verifying an entire B-Tree index level
*/
typedef struct BtreeLevel
{
/* Level number (0 is leaf page level). */
uint32 level;
/* Left most block on level. Scan of level begins here. */
BlockNumber leftmost;
/* Is this level reported as "true" root level by meta page? */
bool istruerootlevel;
} BtreeLevel;
PG_FUNCTION_INFO_V1(bt_index_check);
PG_FUNCTION_INFO_V1(bt_index_parent_check);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
static void bt_index_check_internal(Oid indrelid, bool parentcheck,
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bool heapallindexed, bool rootdescend);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static inline void btree_index_checkable(Relation rel);
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
static inline bool btree_index_mainfork_expected(Relation rel);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
static void bt_check_every_level(Relation rel, Relation heaprel,
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bool heapkeyspace, bool readonly, bool heapallindexed,
bool rootdescend);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static BtreeLevel bt_check_level_from_leftmost(BtreeCheckState *state,
BtreeLevel level);
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
static void bt_recheck_sibling_links(BtreeCheckState *state,
BlockNumber btpo_prev_from_target,
BlockNumber leftcurrent);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static void bt_target_page_check(BtreeCheckState *state);
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
static BTScanInsert bt_right_page_check_scankey(BtreeCheckState *state);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
static void bt_child_check(BtreeCheckState *state, BTScanInsert targetkey,
OffsetNumber downlinkoffnum);
static void bt_child_highkey_check(BtreeCheckState *state,
OffsetNumber target_downlinkoffnum,
Page loaded_child,
uint32 target_level);
static void bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
Harmonize parameter names in contrib code. Make sure that function declarations use names that exactly match the corresponding names from function definitions in contrib code. Like other recent commits that cleaned up function parameter names, this commit was written with help from clang-tidy. Author: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/CAH2-WznJt9CMM9KJTMjJh_zbL5hD9oX44qdJ4aqZtjFi-zA3Tg@mail.gmail.com
2022-09-22 22:59:20 +02:00
BlockNumber blkno, Page page);
Pass ItemPointer not HeapTuple to IndexBuildCallback. Not all AMs use HeapTuples internally, making it inconvenient to pass a HeapTuple. As the index callbacks really only need the TID, not the full tuple, modify callback to only take ItemPointer. Author: Ashwin Agrawal Reviewed-By: Andres Freund Discussion: https://postgr.es/m/CALfoeis6=8ehuR=VNtHvj3z16cYfCwPdTcpaxU+sfSUJ5QgR3g@mail.gmail.com
2019-11-08 09:44:52 +01:00
static void bt_tuple_present_callback(Relation index, ItemPointer tid,
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Datum *values, bool *isnull,
bool tupleIsAlive, void *checkstate);
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
static IndexTuple bt_normalize_tuple(BtreeCheckState *state,
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
IndexTuple itup);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
static inline IndexTuple bt_posting_plain_tuple(IndexTuple itup, int n);
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
static bool bt_rootdescend(BtreeCheckState *state, IndexTuple itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static inline bool offset_is_negative_infinity(BTPageOpaque opaque,
OffsetNumber offset);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
static inline bool invariant_l_offset(BtreeCheckState *state, BTScanInsert key,
OffsetNumber upperbound);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static inline bool invariant_leq_offset(BtreeCheckState *state,
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
BTScanInsert key,
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
OffsetNumber upperbound);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
static inline bool invariant_g_offset(BtreeCheckState *state, BTScanInsert key,
OffsetNumber lowerbound);
static inline bool invariant_l_nontarget_offset(BtreeCheckState *state,
BTScanInsert key,
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
BlockNumber nontargetblock,
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Page nontarget,
OffsetNumber upperbound);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
static Page palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
static inline BTScanInsert bt_mkscankey_pivotsearch(Relation rel,
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
Relation heaprel,
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
IndexTuple itup);
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
static ItemId PageGetItemIdCareful(BtreeCheckState *state, BlockNumber block,
Page page, OffsetNumber offset);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
static inline ItemPointer BTreeTupleGetHeapTIDCareful(BtreeCheckState *state,
IndexTuple itup, bool nonpivot);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
static inline ItemPointer BTreeTupleGetPointsToTID(IndexTuple itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* bt_index_check(index regclass, heapallindexed boolean)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Verify integrity of B-Tree index.
*
* Acquires AccessShareLock on heap & index relations. Does not consider
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* invariants that exist between parent/child pages. Optionally verifies
* that heap does not contain any unindexed or incorrectly indexed tuples.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
Datum
bt_index_check(PG_FUNCTION_ARGS)
{
Oid indrelid = PG_GETARG_OID(0);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
bool heapallindexed = false;
if (PG_NARGS() == 2)
heapallindexed = PG_GETARG_BOOL(1);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bt_index_check_internal(indrelid, false, heapallindexed, false);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
PG_RETURN_VOID();
}
/*
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* bt_index_parent_check(index regclass, heapallindexed boolean)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Verify integrity of B-Tree index.
*
* Acquires ShareLock on heap & index relations. Verifies that downlinks in
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* parent pages are valid lower bounds on child pages. Optionally verifies
* that heap does not contain any unindexed or incorrectly indexed tuples.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
Datum
bt_index_parent_check(PG_FUNCTION_ARGS)
{
Oid indrelid = PG_GETARG_OID(0);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
bool heapallindexed = false;
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bool rootdescend = false;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
if (PG_NARGS() >= 2)
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
heapallindexed = PG_GETARG_BOOL(1);
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
if (PG_NARGS() == 3)
rootdescend = PG_GETARG_BOOL(2);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bt_index_check_internal(indrelid, true, heapallindexed, rootdescend);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
PG_RETURN_VOID();
}
/*
* Helper for bt_index_[parent_]check, coordinating the bulk of the work.
*/
static void
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bt_index_check_internal(Oid indrelid, bool parentcheck, bool heapallindexed,
bool rootdescend)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
Oid heapid;
Relation indrel;
Relation heaprel;
LOCKMODE lockmode;
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
Oid save_userid;
int save_sec_context;
int save_nestlevel;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (parentcheck)
lockmode = ShareLock;
else
lockmode = AccessShareLock;
/*
* We must lock table before index to avoid deadlocks. However, if the
* passed indrelid isn't an index then IndexGetRelation() will fail.
* Rather than emitting a not-very-helpful error message, postpone
* complaining, expecting that the is-it-an-index test below will fail.
*
* In hot standby mode this will raise an error when parentcheck is true.
*/
heapid = IndexGetRelation(indrelid, true);
if (OidIsValid(heapid))
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
{
Replace uses of heap_open et al with the corresponding table_* function. Author: Andres Freund Discussion: https://postgr.es/m/20190111000539.xbv7s6w7ilcvm7dp@alap3.anarazel.de
2019-01-21 19:32:19 +01:00
heaprel = table_open(heapid, lockmode);
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
/*
* Switch to the table owner's userid, so that any index functions are
* run as that user. Also lock down security-restricted operations
* and arrange to make GUC variable changes local to this command.
*/
GetUserIdAndSecContext(&save_userid, &save_sec_context);
SetUserIdAndSecContext(heaprel->rd_rel->relowner,
save_sec_context | SECURITY_RESTRICTED_OPERATION);
save_nestlevel = NewGUCNestLevel();
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
else
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
{
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
heaprel = NULL;
Silence compiler warnings from some older compilers. Since a117cebd6, some older gcc versions issue "variable may be used uninitialized in this function" complaints for brin_summarize_range. Silence that using the same coding pattern as in bt_index_check_internal; arguably, a117cebd6 had too narrow a view of which compilers might give trouble. Nathan Bossart and Tom Lane. Back-patch as the previous commit was. Discussion: https://postgr.es/m/20220601163537.GA2331988@nathanxps13
2022-06-01 23:21:45 +02:00
/* Set these just to suppress "uninitialized variable" warnings */
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
save_userid = InvalidOid;
save_sec_context = -1;
save_nestlevel = -1;
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Open the target index relations separately (like relation_openrv(), but
* with heap relation locked first to prevent deadlocking). In hot
* standby mode this will raise an error when parentcheck is true.
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
*
* There is no need for the usual indcheckxmin usability horizon test
* here, even in the heapallindexed case, because index undergoing
* verification only needs to have entries for a new transaction snapshot.
* (If this is a parentcheck verification, there is no question about
* committed or recently dead heap tuples lacking index entries due to
* concurrent activity.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
indrel = index_open(indrelid, lockmode);
/*
* Since we did the IndexGetRelation call above without any lock, it's
* barely possible that a race against an index drop/recreation could have
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* netted us the wrong table.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
if (heaprel == NULL || heapid != IndexGetRelation(indrelid, false))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_TABLE),
Improve quoting in some error messages
2021-04-14 09:10:57 +02:00
errmsg("could not open parent table of index \"%s\"",
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
RelationGetRelationName(indrel))));
/* Relation suitable for checking as B-Tree? */
btree_index_checkable(indrel);
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
if (btree_index_mainfork_expected(indrel))
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
bool heapkeyspace,
allequalimage;
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
Replace RelationOpenSmgr() with RelationGetSmgr(). The idea behind this patch is to design out bugs like the one fixed by commit 9d523119f. Previously, once one did RelationOpenSmgr(rel), it was considered okay to access rel->rd_smgr directly for some not-very-clear interval. But since that pointer will be cleared by relcache flushes, we had bugs arising from overreliance on a previous RelationOpenSmgr call still being effective. Now, very little code except that in rel.h and relcache.c should ever touch the rd_smgr field directly. The normal coding rule is to use RelationGetSmgr(rel) and not expect the result to be valid for longer than one smgr function call. There are a couple of places where using the function every single time seemed like overkill, but they are now annotated with large warning comments. Amul Sul, after an idea of mine. Discussion: https://postgr.es/m/CANiYTQsU7yMFpQYnv=BrcRVqK_3U3mtAzAsJCaqtzsDHfsUbdQ@mail.gmail.com
2021-07-12 23:01:29 +02:00
if (!smgrexists(RelationGetSmgr(indrel), MAIN_FORKNUM))
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" lacks a main relation fork",
RelationGetRelationName(indrel))));
amcheck: Sanitize metapage's allequalimage field. This will be helpful if it ever proves necessary to revoke an opclass's support for deduplication. Backpatch: 13-, where nbtree deduplication was introduced.
2020-08-07 00:25:49 +02:00
/* Extract metadata from metapage, and sanitize it in passing */
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
_bt_metaversion(indrel, heaprel, &heapkeyspace, &allequalimage);
amcheck: Sanitize metapage's allequalimage field. This will be helpful if it ever proves necessary to revoke an opclass's support for deduplication. Backpatch: 13-, where nbtree deduplication was introduced.
2020-08-07 00:25:49 +02:00
if (allequalimage && !heapkeyspace)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" metapage has equalimage field set on unsupported nbtree version",
RelationGetRelationName(indrel))));
if (allequalimage && !_bt_allequalimage(indrel, false))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" metapage incorrectly indicates that deduplication is safe",
RelationGetRelationName(indrel))));
/* Check index, possibly against table it is an index on */
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
bt_check_every_level(indrel, heaprel, heapkeyspace, parentcheck,
heapallindexed, rootdescend);
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Make relation-enumerating operations be security-restricted operations. When a feature enumerates relations and runs functions associated with all found relations, the feature's user shall not need to trust every user having permission to create objects. BRIN-specific functionality in autovacuum neglected to account for this, as did pg_amcheck and CLUSTER. An attacker having permission to create non-temp objects in at least one schema could execute arbitrary SQL functions under the identity of the bootstrap superuser. CREATE INDEX (not a relation-enumerating operation) and REINDEX protected themselves too late. This change extends to the non-enumerating amcheck interface. Back-patch to v10 (all supported versions). Sergey Shinderuk, reviewed (in earlier versions) by Alexander Lakhin. Reported by Alexander Lakhin. Security: CVE-2022-1552
2022-05-09 17:35:08 +02:00
/* Roll back any GUC changes executed by index functions */
AtEOXact_GUC(false, save_nestlevel);
/* Restore userid and security context */
SetUserIdAndSecContext(save_userid, save_sec_context);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Release locks early. That's ok here because nothing in the called
* routines will trigger shared cache invalidations to be sent, so we can
* relax the usual pattern of only releasing locks after commit.
*/
index_close(indrel, lockmode);
if (heaprel)
Replace uses of heap_open et al with the corresponding table_* function. Author: Andres Freund Discussion: https://postgr.es/m/20190111000539.xbv7s6w7ilcvm7dp@alap3.anarazel.de
2019-01-21 19:32:19 +01:00
table_close(heaprel, lockmode);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
* Basic checks about the suitability of a relation for checking as a B-Tree
* index.
*
* NB: Intentionally not checking permissions, the function is normally not
* callable by non-superusers. If granted, it's useful to be able to check a
* whole cluster.
*/
static inline void
btree_index_checkable(Relation rel)
{
if (rel->rd_rel->relkind != RELKIND_INDEX ||
rel->rd_rel->relam != BTREE_AM_OID)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("only B-Tree indexes are supported as targets for verification"),
errdetail("Relation \"%s\" is not a B-Tree index.",
RelationGetRelationName(rel))));
if (RELATION_IS_OTHER_TEMP(rel))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot access temporary tables of other sessions"),
errdetail("Index \"%s\" is associated with temporary relation.",
RelationGetRelationName(rel))));
Remove obsolete IndexIs* macros Remove IndexIsValid(), IndexIsReady(), IndexIsLive() in favor of accessing the index structure directly. These macros haven't been used consistently, and the original reason of maintaining source compatibility with PostgreSQL 9.2 is gone. Discussion: https://www.postgresql.org/message-id/flat/d419147c-09d4-6196-5d9d-0234b230880a%402ndquadrant.com
2018-12-27 10:07:46 +01:00
if (!rel->rd_index->indisvalid)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot check index \"%s\"",
RelationGetRelationName(rel)),
Fix some errhint and errdetail strings missing a period As per the error message style guide of the documentation, those should be full sentences. Author: Daniel Gustafsson Reviewed-by: Michael Paquier, Álvaro Herrera Discussion: https://1E8D49B4-16BC-4420-B4ED-58501D9E076B@yesql.se
2018-12-06 23:47:42 +01:00
errdetail("Index is not valid.")));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
/*
* Check if B-Tree index relation should have a file for its main relation
* fork. Verification uses this to skip unlogged indexes when in hot standby
amcheck: Skip unlogged relations in Hot Standby. Have verify_heapam.c treat unlogged relations as if they were simply empty when in Hot Standby mode. This brings it in line with verify_nbtree.c, which has handled unlogged relations in the same way since bugfix commit 6754fe65a4. This was an oversight in commit 866e24d47d, which extended contrib/amcheck to check heap relations. In passing, lower the verbosity used when reporting that a relation has been skipped like this, from NOTICE to DEBUG1. This is appropriate because the skipping behavior is only an implementation detail, needed to work around the fact that unlogged tables don't have smgr-level storage for their main fork when in Hot Standby mode. Affected unlogged relations should be considered "trivially verified", not skipped over. They are verified in the same sense that a totally empty relation can be verified. This behavior seems least surprising overall, since unlogged relations on a replica will initially be empty if and when the replica is promoted and Hot Standby ends. Author: Mark Dilger <mark.dilger@enterprisedb.com> Reviewed-By: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/CAH2-Wzk_pukOFY7JmdiFLsrz+Pd3V8OwgC1TH2Vd5BH5ZgK4bA@mail.gmail.com Backpatch: 14-, where heapam verification was introduced.
2021-10-12 02:21:48 +02:00
* mode, where there is simply nothing to verify. We behave as if the
* relation is empty.
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
*
* NB: Caller should call btree_index_checkable() before calling here.
*/
static inline bool
btree_index_mainfork_expected(Relation rel)
{
if (rel->rd_rel->relpersistence != RELPERSISTENCE_UNLOGGED ||
!RecoveryInProgress())
return true;
amcheck: Skip unlogged relations in Hot Standby. Have verify_heapam.c treat unlogged relations as if they were simply empty when in Hot Standby mode. This brings it in line with verify_nbtree.c, which has handled unlogged relations in the same way since bugfix commit 6754fe65a4. This was an oversight in commit 866e24d47d, which extended contrib/amcheck to check heap relations. In passing, lower the verbosity used when reporting that a relation has been skipped like this, from NOTICE to DEBUG1. This is appropriate because the skipping behavior is only an implementation detail, needed to work around the fact that unlogged tables don't have smgr-level storage for their main fork when in Hot Standby mode. Affected unlogged relations should be considered "trivially verified", not skipped over. They are verified in the same sense that a totally empty relation can be verified. This behavior seems least surprising overall, since unlogged relations on a replica will initially be empty if and when the replica is promoted and Hot Standby ends. Author: Mark Dilger <mark.dilger@enterprisedb.com> Reviewed-By: Peter Geoghegan <pg@bowt.ie> Discussion: https://postgr.es/m/CAH2-Wzk_pukOFY7JmdiFLsrz+Pd3V8OwgC1TH2Vd5BH5ZgK4bA@mail.gmail.com Backpatch: 14-, where heapam verification was introduced.
2021-10-12 02:21:48 +02:00
ereport(DEBUG1,
amcheck: Skip unlogged relations during recovery. contrib/amcheck failed to consider the possibility that unlogged relations will not have any main relation fork files when running in hot standby mode. This led to low-level "can't happen" errors that complain about the absence of a relfilenode file. To fix, simply skip verification of unlogged index relations during recovery. In passing, add a direct check for the presence of a main fork just before verification proper begins, so that we cleanly verify the presence of the main relation fork file. Author: Andrey Borodin, Peter Geoghegan Reported-By: Andrey Borodin Diagnosed-By: Andrey Borodin Discussion: https://postgr.es/m/DA9B33AC-53CB-4643-96D4-7A0BBC037FA1@yandex-team.ru Backpatch: 10-, where amcheck was introduced.
2019-08-13 00:21:32 +02:00
(errcode(ERRCODE_READ_ONLY_SQL_TRANSACTION),
errmsg("cannot verify unlogged index \"%s\" during recovery, skipping",
RelationGetRelationName(rel))));
return false;
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Main entry point for B-Tree SQL-callable functions. Walks the B-Tree in
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* logical order, verifying invariants as it goes. Optionally, verification
* checks if the heap relation contains any tuples that are not represented in
* the index but should be.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* It is the caller's responsibility to acquire appropriate heavyweight lock on
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* the index relation, and advise us if extra checks are safe when a ShareLock
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* is held. (A lock of the same type must also have been acquired on the heap
* relation.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* A ShareLock is generally assumed to prevent any kind of physical
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* modification to the index structure, including modifications that VACUUM may
* make. This does not include setting of the LP_DEAD bit by concurrent index
* scans, although that is just metadata that is not able to directly affect
* any check performed here. Any concurrent process that might act on the
* LP_DEAD bit being set (recycle space) requires a heavyweight lock that
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* cannot be held while we hold a ShareLock. (Besides, even if that could
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* happen, the ad-hoc recycling when a page might otherwise split is performed
* per-page, and requires an exclusive buffer lock, which wouldn't cause us
* trouble. _bt_delitems_vacuum() may only delete leaf items, and so the extra
* parent/child check cannot be affected.)
*/
static void
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
bt_check_every_level(Relation rel, Relation heaprel, bool heapkeyspace,
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
bool readonly, bool heapallindexed, bool rootdescend)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
BtreeCheckState *state;
Page metapage;
BTMetaPageData *metad;
uint32 previouslevel;
BtreeLevel current;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Snapshot snapshot = SnapshotAny;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Add contrib/amcheck debug message. Add a DEBUG1 message indicating that verification of the index structure is underway. Also reduce the severity level of the existing "tree level" debug message to DEBUG1. It should never have been made DEBUG2. Any B-Tree index with more than a couple of levels will generally also have so many pages that the per-page DEBUG2 messages will become completely unmanageable. In passing, add a new "Tip" to the docs that advises users that run into corruption that the debug messages might provide useful additional context.
2020-04-11 02:44:08 +02:00
if (!readonly)
elog(DEBUG1, "verifying consistency of tree structure for index \"%s\"",
RelationGetRelationName(rel));
else
elog(DEBUG1, "verifying consistency of tree structure for index \"%s\" with cross-level checks",
RelationGetRelationName(rel));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
snapshot scalability: Don't compute global horizons while building snapshots. To make GetSnapshotData() more scalable, it cannot not look at at each proc's xmin: While snapshot contents do not need to change whenever a read-only transaction commits or a snapshot is released, a proc's xmin is modified in those cases. The frequency of xmin modifications leads to, particularly on higher core count systems, many cache misses inside GetSnapshotData(), despite the data underlying a snapshot not changing. That is the most significant source of GetSnapshotData() scaling poorly on larger systems. Without accessing xmins, GetSnapshotData() cannot calculate accurate horizons / thresholds as it has so far. But we don't really have to: The horizons don't actually change that much between GetSnapshotData() calls. Nor are the horizons actually used every time a snapshot is built. The trick this commit introduces is to delay computation of accurate horizons until there use and using horizon boundaries to determine whether accurate horizons need to be computed. The use of RecentGlobal[Data]Xmin to decide whether a row version could be removed has been replaces with new GlobalVisTest* functions. These use two thresholds to determine whether a row can be pruned: 1) definitely_needed, indicating that rows deleted by XIDs >= definitely_needed are definitely still visible. 2) maybe_needed, indicating that rows deleted by XIDs < maybe_needed can definitely be removed GetSnapshotData() updates definitely_needed to be the xmin of the computed snapshot. When testing whether a row can be removed (with GlobalVisTestIsRemovableXid()) and the tested XID falls in between the two (i.e. XID >= maybe_needed && XID < definitely_needed) the boundaries can be recomputed to be more accurate. As it is not cheap to compute accurate boundaries, we limit the number of times that happens in short succession. As the boundaries used by GlobalVisTestIsRemovableXid() are never reset (with maybe_needed updated by GetSnapshotData()), it is likely that further test can benefit from an earlier computation of accurate horizons. To avoid regressing performance when old_snapshot_threshold is set (as that requires an accurate horizon to be computed), heap_page_prune_opt() doesn't unconditionally call TransactionIdLimitedForOldSnapshots() anymore. Both the computation of the limited horizon, and the triggering of errors (with SetOldSnapshotThresholdTimestamp()) is now only done when necessary to remove tuples. This commit just removes the accesses to PGXACT->xmin from GetSnapshotData(), but other members of PGXACT residing in the same cache line are accessed. Therefore this in itself does not result in a significant improvement. Subsequent commits will take advantage of the fact that GetSnapshotData() now does not need to access xmins anymore. Note: This contains a workaround in heap_page_prune_opt() to keep the snapshot_too_old tests working. While that workaround is ugly, the tests currently are not meaningful, and it seems best to address them separately. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Robert Haas <robertmhaas@gmail.com> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: David Rowley <dgrowleyml@gmail.com> Discussion: https://postgr.es/m/20200301083601.ews6hz5dduc3w2se@alap3.anarazel.de
2020-08-13 01:03:49 +02:00
* This assertion matches the one in index_getnext_tid(). See page
* recycling/"visible to everyone" notes in nbtree README.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
snapshot scalability: Don't compute global horizons while building snapshots. To make GetSnapshotData() more scalable, it cannot not look at at each proc's xmin: While snapshot contents do not need to change whenever a read-only transaction commits or a snapshot is released, a proc's xmin is modified in those cases. The frequency of xmin modifications leads to, particularly on higher core count systems, many cache misses inside GetSnapshotData(), despite the data underlying a snapshot not changing. That is the most significant source of GetSnapshotData() scaling poorly on larger systems. Without accessing xmins, GetSnapshotData() cannot calculate accurate horizons / thresholds as it has so far. But we don't really have to: The horizons don't actually change that much between GetSnapshotData() calls. Nor are the horizons actually used every time a snapshot is built. The trick this commit introduces is to delay computation of accurate horizons until there use and using horizon boundaries to determine whether accurate horizons need to be computed. The use of RecentGlobal[Data]Xmin to decide whether a row version could be removed has been replaces with new GlobalVisTest* functions. These use two thresholds to determine whether a row can be pruned: 1) definitely_needed, indicating that rows deleted by XIDs >= definitely_needed are definitely still visible. 2) maybe_needed, indicating that rows deleted by XIDs < maybe_needed can definitely be removed GetSnapshotData() updates definitely_needed to be the xmin of the computed snapshot. When testing whether a row can be removed (with GlobalVisTestIsRemovableXid()) and the tested XID falls in between the two (i.e. XID >= maybe_needed && XID < definitely_needed) the boundaries can be recomputed to be more accurate. As it is not cheap to compute accurate boundaries, we limit the number of times that happens in short succession. As the boundaries used by GlobalVisTestIsRemovableXid() are never reset (with maybe_needed updated by GetSnapshotData()), it is likely that further test can benefit from an earlier computation of accurate horizons. To avoid regressing performance when old_snapshot_threshold is set (as that requires an accurate horizon to be computed), heap_page_prune_opt() doesn't unconditionally call TransactionIdLimitedForOldSnapshots() anymore. Both the computation of the limited horizon, and the triggering of errors (with SetOldSnapshotThresholdTimestamp()) is now only done when necessary to remove tuples. This commit just removes the accesses to PGXACT->xmin from GetSnapshotData(), but other members of PGXACT residing in the same cache line are accessed. Therefore this in itself does not result in a significant improvement. Subsequent commits will take advantage of the fact that GetSnapshotData() now does not need to access xmins anymore. Note: This contains a workaround in heap_page_prune_opt() to keep the snapshot_too_old tests working. While that workaround is ugly, the tests currently are not meaningful, and it seems best to address them separately. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Robert Haas <robertmhaas@gmail.com> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: David Rowley <dgrowleyml@gmail.com> Discussion: https://postgr.es/m/20200301083601.ews6hz5dduc3w2se@alap3.anarazel.de
2020-08-13 01:03:49 +02:00
Assert(TransactionIdIsValid(RecentXmin));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Initialize state for entire verification operation
*/
Add amcheck missing downlink tests. Also use palloc0() for main amcheck state, and adjust a few comments. Somehow I pushed old version of patch in commit 4eaf7eaccb291f5d32d9f05284bfca9c5744de9a, so commit the difference. Peter Geoghegan
2018-04-25 21:05:53 +02:00
state = palloc0(sizeof(BtreeCheckState));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
state->rel = rel;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
state->heaprel = heaprel;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
state->heapkeyspace = heapkeyspace;
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
state->readonly = readonly;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
state->heapallindexed = heapallindexed;
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
state->rootdescend = rootdescend;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
if (state->heapallindexed)
{
Don't rely on estimates for amcheck Bloom filters. Solely relying on a relation's reltuples/relpages estimate to size the Bloom filters used by amcheck verification makes verification less effective when the estimates are very stale. In extreme cases, verification options that use Bloom filters internally could be totally ineffective, without users receiving any clear indication that certain types of corruption might easily be missed. To fix, use RelationGetNumberOfBlocks() instead of relpages to size the downlink block Bloom filter. Use the same RelationGetNumberOfBlocks() value to derive a minimum size for the heapallindexed Bloom filter, rather than completely trusting reltuples. Verification will still be reasonably effective when the projected/estimated number of Bloom filter elements is at least 1/5 of the final number of elements, which is assured by the new sizing logic. Reported-By: Alexander Korotkov Discussion: https://postgr.es/m/CAH2-Wzk0ke2J42KrNYBKu0Xovjy-sU5ub7PWjgpbsKdAQcL4OA@mail.gmail.com Backpatch: 11-, where downlink/heapallindexed verification were added.
2019-07-20 20:11:55 +02:00
int64 total_pages;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
int64 total_elems;
uint64 seed;
Don't rely on estimates for amcheck Bloom filters. Solely relying on a relation's reltuples/relpages estimate to size the Bloom filters used by amcheck verification makes verification less effective when the estimates are very stale. In extreme cases, verification options that use Bloom filters internally could be totally ineffective, without users receiving any clear indication that certain types of corruption might easily be missed. To fix, use RelationGetNumberOfBlocks() instead of relpages to size the downlink block Bloom filter. Use the same RelationGetNumberOfBlocks() value to derive a minimum size for the heapallindexed Bloom filter, rather than completely trusting reltuples. Verification will still be reasonably effective when the projected/estimated number of Bloom filter elements is at least 1/5 of the final number of elements, which is assured by the new sizing logic. Reported-By: Alexander Korotkov Discussion: https://postgr.es/m/CAH2-Wzk0ke2J42KrNYBKu0Xovjy-sU5ub7PWjgpbsKdAQcL4OA@mail.gmail.com Backpatch: 11-, where downlink/heapallindexed verification were added.
2019-07-20 20:11:55 +02:00
/*
* Size Bloom filter based on estimated number of tuples in index,
* while conservatively assuming that each block must contain at least
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
* MaxTIDsPerBTreePage / 3 "plain" tuples -- see
* bt_posting_plain_tuple() for definition, and details of how posting
* list tuples are handled.
Don't rely on estimates for amcheck Bloom filters. Solely relying on a relation's reltuples/relpages estimate to size the Bloom filters used by amcheck verification makes verification less effective when the estimates are very stale. In extreme cases, verification options that use Bloom filters internally could be totally ineffective, without users receiving any clear indication that certain types of corruption might easily be missed. To fix, use RelationGetNumberOfBlocks() instead of relpages to size the downlink block Bloom filter. Use the same RelationGetNumberOfBlocks() value to derive a minimum size for the heapallindexed Bloom filter, rather than completely trusting reltuples. Verification will still be reasonably effective when the projected/estimated number of Bloom filter elements is at least 1/5 of the final number of elements, which is assured by the new sizing logic. Reported-By: Alexander Korotkov Discussion: https://postgr.es/m/CAH2-Wzk0ke2J42KrNYBKu0Xovjy-sU5ub7PWjgpbsKdAQcL4OA@mail.gmail.com Backpatch: 11-, where downlink/heapallindexed verification were added.
2019-07-20 20:11:55 +02:00
*/
total_pages = RelationGetNumberOfBlocks(rel);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
total_elems = Max(total_pages * (MaxTIDsPerBTreePage / 3),
Don't rely on estimates for amcheck Bloom filters. Solely relying on a relation's reltuples/relpages estimate to size the Bloom filters used by amcheck verification makes verification less effective when the estimates are very stale. In extreme cases, verification options that use Bloom filters internally could be totally ineffective, without users receiving any clear indication that certain types of corruption might easily be missed. To fix, use RelationGetNumberOfBlocks() instead of relpages to size the downlink block Bloom filter. Use the same RelationGetNumberOfBlocks() value to derive a minimum size for the heapallindexed Bloom filter, rather than completely trusting reltuples. Verification will still be reasonably effective when the projected/estimated number of Bloom filter elements is at least 1/5 of the final number of elements, which is assured by the new sizing logic. Reported-By: Alexander Korotkov Discussion: https://postgr.es/m/CAH2-Wzk0ke2J42KrNYBKu0Xovjy-sU5ub7PWjgpbsKdAQcL4OA@mail.gmail.com Backpatch: 11-, where downlink/heapallindexed verification were added.
2019-07-20 20:11:55 +02:00
(int64) state->rel->rd_rel->reltuples);
Replace random(), pg_erand48(), etc with a better PRNG API and algorithm. Standardize on xoroshiro128** as our basic PRNG algorithm, eliminating a bunch of platform dependencies as well as fundamentally-obsolete PRNG code. In addition, this API replacement will ease replacing the algorithm again in future, should that become necessary. xoroshiro128** is a few percent slower than the drand48 family, but it can produce full-width 64-bit random values not only 48-bit, and it should be much more trustworthy. It's likely to be noticeably faster than the platform's random(), depending on which platform you are thinking about; and we can have non-global state vectors easily, unlike with random(). It is not cryptographically strong, but neither are the functions it replaces. Fabien Coelho, reviewed by Dean Rasheed, Aleksander Alekseev, and myself Discussion: https://postgr.es/m/alpine.DEB.2.22.394.2105241211230.165418@pseudo
2021-11-29 03:32:36 +01:00
/* Generate a random seed to avoid repetition */
seed = pg_prng_uint64(&pg_global_prng_state);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* Create Bloom filter to fingerprint index */
state->filter = bloom_create(total_elems, maintenance_work_mem, seed);
state->heaptuplespresent = 0;
/*
* Register our own snapshot in !readonly case, rather than asking
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* table_index_build_scan() to do this for us later. This needs to
* happen before index fingerprinting begins, so we can later be
* certain that index fingerprinting should have reached all tuples
* returned by table_index_build_scan().
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
*/
if (!state->readonly)
{
snapshot = RegisterSnapshot(GetTransactionSnapshot());
/*
* GetTransactionSnapshot() always acquires a new MVCC snapshot in
* READ COMMITTED mode. A new snapshot is guaranteed to have all
* the entries it requires in the index.
*
* We must defend against the possibility that an old xact
* snapshot was returned at higher isolation levels when that
* snapshot is not safe for index scans of the target index. This
* is possible when the snapshot sees tuples that are before the
* index's indcheckxmin horizon. Throwing an error here should be
* very rare. It doesn't seem worth using a secondary snapshot to
* avoid this.
*/
if (IsolationUsesXactSnapshot() && rel->rd_index->indcheckxmin &&
!TransactionIdPrecedes(HeapTupleHeaderGetXmin(rel->rd_indextuple->t_data),
snapshot->xmin))
ereport(ERROR,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("index \"%s\" cannot be verified using transaction snapshot",
RelationGetRelationName(rel))));
}
}
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
Assert(!state->rootdescend || state->readonly);
if (state->rootdescend && !state->heapkeyspace)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot verify that tuples from index \"%s\" can each be found by an independent index search",
RelationGetRelationName(rel)),
errhint("Only B-Tree version 4 indexes support rootdescend verification.")));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Create context for page */
state->targetcontext = AllocSetContextCreate(CurrentMemoryContext,
"amcheck context",
Rethink MemoryContext creation to improve performance. This patch makes a number of interrelated changes to reduce the overhead involved in creating/deleting memory contexts. The key ideas are: * Include the AllocSetContext header of an aset.c context in its first malloc request, rather than allocating it separately in TopMemoryContext. This means that we now always create an initial or "keeper" block in an aset, even if it never receives any allocation requests. * Create freelists in which we can save and recycle recently-destroyed asets (this idea is due to Robert Haas). * In the common case where the name of a context is a constant string, just store a pointer to it in the context header, rather than copying the string. The first change eliminates a palloc/pfree cycle per context, and also avoids bloat in TopMemoryContext, at the price that creating a context now involves a malloc/free cycle even if the context never receives any allocations. That would be a loser for some common usage patterns, but recycling short-lived contexts via the freelist eliminates that pain. Avoiding copying constant strings not only saves strlen() and strcpy() overhead, but is an essential part of the freelist optimization because it makes the context header size constant. Currently we make no attempt to use the freelist for contexts with non-constant names. (Perhaps someday we'll need to think harder about that, but in current usage, most contexts with custom names are long-lived anyway.) The freelist management in this initial commit is pretty simplistic, and we might want to refine it later --- but in common workloads that will never matter because the freelists will never get full anyway. To create a context with a non-constant name, one is now required to call AllocSetContextCreateExtended and specify the MEMCONTEXT_COPY_NAME option. AllocSetContextCreate becomes a wrapper macro, and it includes a test that will complain about non-string-literal context name parameters on gcc and similar compilers. An unfortunate side effect of making AllocSetContextCreate a macro is that one is now *required* to use the size parameter abstraction macros (ALLOCSET_DEFAULT_SIZES and friends) with it; the pre-9.6 habit of writing out individual size parameters no longer works unless you switch to AllocSetContextCreateExtended. Internally to the memory-context-related modules, the context creation APIs are simplified, removing the rather baroque original design whereby a context-type module called mcxt.c which then called back into the context-type module. That saved a bit of code duplication, but not much, and it prevented context-type modules from exercising control over the allocation of context headers. In passing, I converted the test-and-elog validation of aset size parameters into Asserts to save a few more cycles. The original thought was that callers might compute size parameters on the fly, but in practice nobody does that, so it's useless to expend cycles on checking those numbers in production builds. Also, mark the memory context method-pointer structs "const", just for cleanliness. Discussion: https://postgr.es/m/2264.1512870796@sss.pgh.pa.us
2017-12-13 19:55:12 +01:00
ALLOCSET_DEFAULT_SIZES);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
state->checkstrategy = GetAccessStrategy(BAS_BULKREAD);
/* Get true root block from meta-page */
metapage = palloc_btree_page(state, BTREE_METAPAGE);
metad = BTPageGetMeta(metapage);
/*
* Certain deletion patterns can result in "skinny" B-Tree indexes, where
* the fast root and true root differ.
*
* Start from the true root, not the fast root, unlike conventional index
* scans. This approach is more thorough, and removes the risk of
* following a stale fast root from the meta page.
*/
if (metad->btm_fastroot != metad->btm_root)
ereport(DEBUG1,
(errcode(ERRCODE_NO_DATA),
Improve quoting in some error messages
2021-04-14 09:10:57 +02:00
errmsg_internal("harmless fast root mismatch in index \"%s\"",
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
RelationGetRelationName(rel)),
errdetail_internal("Fast root block %u (level %u) differs from true root block %u (level %u).",
metad->btm_fastroot, metad->btm_fastlevel,
metad->btm_root, metad->btm_level)));
/*
* Starting at the root, verify every level. Move left to right, top to
* bottom. Note that there may be no pages other than the meta page (meta
* page can indicate that root is P_NONE when the index is totally empty).
*/
previouslevel = InvalidBtreeLevel;
current.level = metad->btm_level;
current.leftmost = metad->btm_root;
current.istruerootlevel = true;
while (current.leftmost != P_NONE)
{
/*
* Verify this level, and get left most page for next level down, if
* not at leaf level
*/
current = bt_check_level_from_leftmost(state, current);
if (current.leftmost == InvalidBlockNumber)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" has no valid pages on level below %u or first level",
RelationGetRelationName(rel), previouslevel)));
previouslevel = current.level;
}
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/*
* * Check whether heap contains unindexed/malformed tuples *
*/
if (state->heapallindexed)
{
IndexInfo *indexinfo = BuildIndexInfo(state->rel);
tableam: Add and use scan APIs. Too allow table accesses to be not directly dependent on heap, several new abstractions are needed. Specifically: 1) Heap scans need to be generalized into table scans. Do this by introducing TableScanDesc, which will be the "base class" for individual AMs. This contains the AM independent fields from HeapScanDesc. The previous heap_{beginscan,rescan,endscan} et al. have been replaced with a table_ version. There's no direct replacement for heap_getnext(), as that returned a HeapTuple, which is undesirable for a other AMs. Instead there's table_scan_getnextslot(). But note that heap_getnext() lives on, it's still used widely to access catalog tables. This is achieved by new scan_begin, scan_end, scan_rescan, scan_getnextslot callbacks. 2) The portion of parallel scans that's shared between backends need to be able to do so without the user doing per-AM work. To achieve that new parallelscan_{estimate, initialize, reinitialize} callbacks are introduced, which operate on a new ParallelTableScanDesc, which again can be subclassed by AMs. As it is likely that several AMs are going to be block oriented, block oriented callbacks that can be shared between such AMs are provided and used by heap. table_block_parallelscan_{estimate, intiialize, reinitialize} as callbacks, and table_block_parallelscan_{nextpage, init} for use in AMs. These operate on a ParallelBlockTableScanDesc. 3) Index scans need to be able to access tables to return a tuple, and there needs to be state across individual accesses to the heap to store state like buffers. That's now handled by introducing a sort-of-scan IndexFetchTable, which again is intended to be subclassed by individual AMs (for heap IndexFetchHeap). The relevant callbacks for an AM are index_fetch_{end, begin, reset} to create the necessary state, and index_fetch_tuple to retrieve an indexed tuple. Note that index_fetch_tuple implementations need to be smarter than just blindly fetching the tuples for AMs that have optimizations similar to heap's HOT - the currently alive tuple in the update chain needs to be fetched if appropriate. Similar to table_scan_getnextslot(), it's undesirable to continue to return HeapTuples. Thus index_fetch_heap (might want to rename that later) now accepts a slot as an argument. Core code doesn't have a lot of call sites performing index scans without going through the systable_* API (in contrast to loads of heap_getnext calls and working directly with HeapTuples). Index scans now store the result of a search in IndexScanDesc->xs_heaptid, rather than xs_ctup->t_self. As the target is not generally a HeapTuple anymore that seems cleaner. To be able to sensible adapt code to use the above, two further callbacks have been introduced: a) slot_callbacks returns a TupleTableSlotOps* suitable for creating slots capable of holding a tuple of the AMs type. table_slot_callbacks() and table_slot_create() are based upon that, but have additional logic to deal with views, foreign tables, etc. While this change could have been done separately, nearly all the call sites that needed to be adapted for the rest of this commit also would have been needed to be adapted for table_slot_callbacks(), making separation not worthwhile. b) tuple_satisfies_snapshot checks whether the tuple in a slot is currently visible according to a snapshot. That's required as a few places now don't have a buffer + HeapTuple around, but a slot (which in heap's case internally has that information). Additionally a few infrastructure changes were needed: I) SysScanDesc, as used by systable_{beginscan, getnext} et al. now internally uses a slot to keep track of tuples. While systable_getnext() still returns HeapTuples, and will so for the foreseeable future, the index API (see 1) above) now only deals with slots. The remainder, and largest part, of this commit is then adjusting all scans in postgres to use the new APIs. Author: Andres Freund, Haribabu Kommi, Alvaro Herrera Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de https://postgr.es/m/20160812231527.GA690404@alvherre.pgsql
2019-03-11 20:46:41 +01:00
TableScanDesc scan;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/*
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* Create our own scan for table_index_build_scan(), rather than
* getting it to do so for us. This is required so that we can
* actually use the MVCC snapshot registered earlier in !readonly
* case.
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
*
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* Note that table_index_build_scan() calls heap_endscan() for us.
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
*/
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
scan = table_beginscan_strat(state->heaprel, /* relation */
tableam: Add and use scan APIs. Too allow table accesses to be not directly dependent on heap, several new abstractions are needed. Specifically: 1) Heap scans need to be generalized into table scans. Do this by introducing TableScanDesc, which will be the "base class" for individual AMs. This contains the AM independent fields from HeapScanDesc. The previous heap_{beginscan,rescan,endscan} et al. have been replaced with a table_ version. There's no direct replacement for heap_getnext(), as that returned a HeapTuple, which is undesirable for a other AMs. Instead there's table_scan_getnextslot(). But note that heap_getnext() lives on, it's still used widely to access catalog tables. This is achieved by new scan_begin, scan_end, scan_rescan, scan_getnextslot callbacks. 2) The portion of parallel scans that's shared between backends need to be able to do so without the user doing per-AM work. To achieve that new parallelscan_{estimate, initialize, reinitialize} callbacks are introduced, which operate on a new ParallelTableScanDesc, which again can be subclassed by AMs. As it is likely that several AMs are going to be block oriented, block oriented callbacks that can be shared between such AMs are provided and used by heap. table_block_parallelscan_{estimate, intiialize, reinitialize} as callbacks, and table_block_parallelscan_{nextpage, init} for use in AMs. These operate on a ParallelBlockTableScanDesc. 3) Index scans need to be able to access tables to return a tuple, and there needs to be state across individual accesses to the heap to store state like buffers. That's now handled by introducing a sort-of-scan IndexFetchTable, which again is intended to be subclassed by individual AMs (for heap IndexFetchHeap). The relevant callbacks for an AM are index_fetch_{end, begin, reset} to create the necessary state, and index_fetch_tuple to retrieve an indexed tuple. Note that index_fetch_tuple implementations need to be smarter than just blindly fetching the tuples for AMs that have optimizations similar to heap's HOT - the currently alive tuple in the update chain needs to be fetched if appropriate. Similar to table_scan_getnextslot(), it's undesirable to continue to return HeapTuples. Thus index_fetch_heap (might want to rename that later) now accepts a slot as an argument. Core code doesn't have a lot of call sites performing index scans without going through the systable_* API (in contrast to loads of heap_getnext calls and working directly with HeapTuples). Index scans now store the result of a search in IndexScanDesc->xs_heaptid, rather than xs_ctup->t_self. As the target is not generally a HeapTuple anymore that seems cleaner. To be able to sensible adapt code to use the above, two further callbacks have been introduced: a) slot_callbacks returns a TupleTableSlotOps* suitable for creating slots capable of holding a tuple of the AMs type. table_slot_callbacks() and table_slot_create() are based upon that, but have additional logic to deal with views, foreign tables, etc. While this change could have been done separately, nearly all the call sites that needed to be adapted for the rest of this commit also would have been needed to be adapted for table_slot_callbacks(), making separation not worthwhile. b) tuple_satisfies_snapshot checks whether the tuple in a slot is currently visible according to a snapshot. That's required as a few places now don't have a buffer + HeapTuple around, but a slot (which in heap's case internally has that information). Additionally a few infrastructure changes were needed: I) SysScanDesc, as used by systable_{beginscan, getnext} et al. now internally uses a slot to keep track of tuples. While systable_getnext() still returns HeapTuples, and will so for the foreseeable future, the index API (see 1) above) now only deals with slots. The remainder, and largest part, of this commit is then adjusting all scans in postgres to use the new APIs. Author: Andres Freund, Haribabu Kommi, Alvaro Herrera Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de https://postgr.es/m/20160812231527.GA690404@alvherre.pgsql
2019-03-11 20:46:41 +01:00
snapshot, /* snapshot */
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
0, /* number of keys */
tableam: Add and use scan APIs. Too allow table accesses to be not directly dependent on heap, several new abstractions are needed. Specifically: 1) Heap scans need to be generalized into table scans. Do this by introducing TableScanDesc, which will be the "base class" for individual AMs. This contains the AM independent fields from HeapScanDesc. The previous heap_{beginscan,rescan,endscan} et al. have been replaced with a table_ version. There's no direct replacement for heap_getnext(), as that returned a HeapTuple, which is undesirable for a other AMs. Instead there's table_scan_getnextslot(). But note that heap_getnext() lives on, it's still used widely to access catalog tables. This is achieved by new scan_begin, scan_end, scan_rescan, scan_getnextslot callbacks. 2) The portion of parallel scans that's shared between backends need to be able to do so without the user doing per-AM work. To achieve that new parallelscan_{estimate, initialize, reinitialize} callbacks are introduced, which operate on a new ParallelTableScanDesc, which again can be subclassed by AMs. As it is likely that several AMs are going to be block oriented, block oriented callbacks that can be shared between such AMs are provided and used by heap. table_block_parallelscan_{estimate, intiialize, reinitialize} as callbacks, and table_block_parallelscan_{nextpage, init} for use in AMs. These operate on a ParallelBlockTableScanDesc. 3) Index scans need to be able to access tables to return a tuple, and there needs to be state across individual accesses to the heap to store state like buffers. That's now handled by introducing a sort-of-scan IndexFetchTable, which again is intended to be subclassed by individual AMs (for heap IndexFetchHeap). The relevant callbacks for an AM are index_fetch_{end, begin, reset} to create the necessary state, and index_fetch_tuple to retrieve an indexed tuple. Note that index_fetch_tuple implementations need to be smarter than just blindly fetching the tuples for AMs that have optimizations similar to heap's HOT - the currently alive tuple in the update chain needs to be fetched if appropriate. Similar to table_scan_getnextslot(), it's undesirable to continue to return HeapTuples. Thus index_fetch_heap (might want to rename that later) now accepts a slot as an argument. Core code doesn't have a lot of call sites performing index scans without going through the systable_* API (in contrast to loads of heap_getnext calls and working directly with HeapTuples). Index scans now store the result of a search in IndexScanDesc->xs_heaptid, rather than xs_ctup->t_self. As the target is not generally a HeapTuple anymore that seems cleaner. To be able to sensible adapt code to use the above, two further callbacks have been introduced: a) slot_callbacks returns a TupleTableSlotOps* suitable for creating slots capable of holding a tuple of the AMs type. table_slot_callbacks() and table_slot_create() are based upon that, but have additional logic to deal with views, foreign tables, etc. While this change could have been done separately, nearly all the call sites that needed to be adapted for the rest of this commit also would have been needed to be adapted for table_slot_callbacks(), making separation not worthwhile. b) tuple_satisfies_snapshot checks whether the tuple in a slot is currently visible according to a snapshot. That's required as a few places now don't have a buffer + HeapTuple around, but a slot (which in heap's case internally has that information). Additionally a few infrastructure changes were needed: I) SysScanDesc, as used by systable_{beginscan, getnext} et al. now internally uses a slot to keep track of tuples. While systable_getnext() still returns HeapTuples, and will so for the foreseeable future, the index API (see 1) above) now only deals with slots. The remainder, and largest part, of this commit is then adjusting all scans in postgres to use the new APIs. Author: Andres Freund, Haribabu Kommi, Alvaro Herrera Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de https://postgr.es/m/20160812231527.GA690404@alvherre.pgsql
2019-03-11 20:46:41 +01:00
NULL, /* scan key */
true, /* buffer access strategy OK */
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
true); /* syncscan OK? */
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/*
* Scan will behave as the first scan of a CREATE INDEX CONCURRENTLY
* behaves in !readonly case.
*
* It's okay that we don't actually use the same lock strength for the
* heap relation as any other ii_Concurrent caller would in !readonly
* case. We have no reason to care about a concurrent VACUUM
* operation, since there isn't going to be a second scan of the heap
* that needs to be sure that there was no concurrent recycling of
* TIDs.
*/
indexinfo->ii_Concurrent = !state->readonly;
/*
* Don't wait for uncommitted tuple xact commit/abort when index is a
* unique index on a catalog (or an index used by an exclusion
* constraint). This could otherwise happen in the readonly case.
*/
indexinfo->ii_Unique = false;
indexinfo->ii_ExclusionOps = NULL;
indexinfo->ii_ExclusionProcs = NULL;
indexinfo->ii_ExclusionStrats = NULL;
elog(DEBUG1, "verifying that tuples from index \"%s\" are present in \"%s\"",
RelationGetRelationName(state->rel),
RelationGetRelationName(state->heaprel));
Report progress of CREATE INDEX operations This uses the progress reporting infrastructure added by c16dc1aca5e0, adding support for CREATE INDEX and CREATE INDEX CONCURRENTLY. There are two pieces to this: one is index-AM-agnostic, and the other is AM-specific. The latter is fairly elaborate for btrees, including reportage for parallel index builds and the separate phases that btree index creation uses; other index AMs, which are much simpler in their building procedures, have simplistic reporting only, but that seems sufficient, at least for non-concurrent builds. The index-AM-agnostic part is fairly complete, providing insight into the CONCURRENTLY wait phases as well as block-based progress during the index validation table scan. (The index validation index scan requires patching each AM, which has not been included here.) Reviewers: Rahila Syed, Pavan Deolasee, Tatsuro Yamada Discussion: https://postgr.es/m/20181220220022.mg63bhk26zdpvmcj@alvherre.pgsql
2019-04-02 20:18:08 +02:00
table_index_build_scan(state->heaprel, state->rel, indexinfo, true, false,
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
bt_tuple_present_callback, (void *) state, scan);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
ereport(DEBUG1,
(errmsg_internal("finished verifying presence of " INT64_FORMAT " tuples from table \"%s\" with bitset %.2f%% set",
state->heaptuplespresent, RelationGetRelationName(heaprel),
100.0 * bloom_prop_bits_set(state->filter))));
if (snapshot != SnapshotAny)
UnregisterSnapshot(snapshot);
bloom_free(state->filter);
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Be tidy: */
MemoryContextDelete(state->targetcontext);
}
/*
* Given a left-most block at some level, move right, verifying each page
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* individually (with more verification across pages for "readonly"
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* callers). Caller should pass the true root page as the leftmost initially,
* working their way down by passing what is returned for the last call here
* until level 0 (leaf page level) was reached.
*
* Returns state for next call, if any. This includes left-most block number
* one level lower that should be passed on next level/call, which is set to
* P_NONE on last call here (when leaf level is verified). Level numbers
* follow the nbtree convention: higher levels have higher numbers, because new
* levels are added only due to a root page split. Note that prior to the
* first root page split, the root is also a leaf page, so there is always a
* level 0 (leaf level), and it's always the last level processed.
*
* Note on memory management: State's per-page context is reset here, between
* each call to bt_target_page_check().
*/
static BtreeLevel
bt_check_level_from_leftmost(BtreeCheckState *state, BtreeLevel level)
{
/* State to establish early, concerning entire level */
BTPageOpaque opaque;
MemoryContext oldcontext;
BtreeLevel nextleveldown;
/* Variables for iterating across level using right links */
BlockNumber leftcurrent = P_NONE;
BlockNumber current = level.leftmost;
/* Initialize return state */
nextleveldown.leftmost = InvalidBlockNumber;
nextleveldown.level = InvalidBtreeLevel;
nextleveldown.istruerootlevel = false;
/* Use page-level context for duration of this call */
oldcontext = MemoryContextSwitchTo(state->targetcontext);
Add contrib/amcheck debug message. Add a DEBUG1 message indicating that verification of the index structure is underway. Also reduce the severity level of the existing "tree level" debug message to DEBUG1. It should never have been made DEBUG2. Any B-Tree index with more than a couple of levels will generally also have so many pages that the per-page DEBUG2 messages will become completely unmanageable. In passing, add a new "Tip" to the docs that advises users that run into corruption that the debug messages might provide useful additional context.
2020-04-11 02:44:08 +02:00
elog(DEBUG1, "verifying level %u%s", level.level,
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
level.istruerootlevel ?
" (true root level)" : level.level == 0 ? " (leaf level)" : "");
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
state->prevrightlink = InvalidBlockNumber;
state->previncompletesplit = false;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
do
{
/* Don't rely on CHECK_FOR_INTERRUPTS() calls at lower level */
CHECK_FOR_INTERRUPTS();
/* Initialize state for this iteration */
state->targetblock = current;
state->target = palloc_btree_page(state, state->targetblock);
state->targetlsn = PageGetLSN(state->target);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(state->target);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (P_IGNORE(opaque))
{
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Since there cannot be a concurrent VACUUM operation in readonly
* mode, and since a page has no links within other pages
* (siblings and parent) once it is marked fully deleted, it
* should be impossible to land on a fully deleted page in
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* readonly mode. See bt_child_check() for further details.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* The bt_child_check() P_ISDELETED() check is repeated here so
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
* that pages that are only reachable through sibling links get
* checked.
*/
if (state->readonly && P_ISDELETED(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("downlink or sibling link points to deleted block in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u left block=%u left link from block=%u.",
current, leftcurrent, opaque->btpo_prev)));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (P_RIGHTMOST(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("block %u fell off the end of index \"%s\"",
current, RelationGetRelationName(state->rel))));
else
ereport(DEBUG1,
(errcode(ERRCODE_NO_DATA),
amcheck: Reduce debug message verbosity. Empty sibling pages can occasionally be much more common than any other event that we report on at elevel DEBUG1. Increase the elevel for relevant cases to DEBUG2 to avoid overwhelming the user with relatively insignificant details.
2021-03-16 21:11:17 +01:00
errmsg_internal("block %u of index \"%s\" concurrently deleted",
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
current, RelationGetRelationName(state->rel))));
goto nextpage;
}
else if (nextleveldown.leftmost == InvalidBlockNumber)
{
/*
* A concurrent page split could make the caller supplied leftmost
* block no longer contain the leftmost page, or no longer be the
* true root, but where that isn't possible due to heavyweight
* locking, check that the first valid page meets caller's
* expectations.
*/
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
if (state->readonly)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
if (!P_LEFTMOST(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("block %u is not leftmost in index \"%s\"",
current, RelationGetRelationName(state->rel))));
if (level.istruerootlevel && !P_ISROOT(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("block %u is not true root in index \"%s\"",
current, RelationGetRelationName(state->rel))));
}
/*
* Before beginning any non-trivial examination of level, prepare
* state for next bt_check_level_from_leftmost() invocation for
* the next level for the next level down (if any).
*
* There should be at least one non-ignorable page per level,
* unless this is the leaf level, which is assumed by caller to be
* final level.
*/
if (!P_ISLEAF(opaque))
{
IndexTuple itup;
ItemId itemid;
/* Internal page -- downlink gets leftmost on next level */
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target,
P_FIRSTDATAKEY(opaque));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
itup = (IndexTuple) PageGetItem(state->target, itemid);
Rename nbtree tuple macros. Rename two function-style macros, removing the word "inner". This makes things more consistent.
2019-12-17 02:49:45 +01:00
nextleveldown.leftmost = BTreeTupleGetDownLink(itup);
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
nextleveldown.level = opaque->btpo_level - 1;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
else
{
/*
* Leaf page -- final level caller must process.
*
* Note that this could also be the root page, if there has
* been no root page split yet.
*/
nextleveldown.leftmost = P_NONE;
nextleveldown.level = InvalidBtreeLevel;
}
/*
* Finished setting up state for this call/level. Control will
* never end up back here in any future loop iteration for this
* level.
*/
}
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
/* Sibling links should be in mutual agreement */
if (opaque->btpo_prev != leftcurrent)
bt_recheck_sibling_links(state, opaque->btpo_prev, leftcurrent);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
/* Check level */
if (level.level != opaque->btpo_level)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("leftmost down link for level points to block in index \"%s\" whose level is not one level down",
RelationGetRelationName(state->rel)),
errdetail_internal("Block pointed to=%u expected level=%u level in pointed to block=%u.",
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
current, level.level, opaque->btpo_level)));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* Verify invariants for page */
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
bt_target_page_check(state);
nextpage:
/* Try to detect circular links */
if (current == leftcurrent || current == opaque->btpo_prev)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("circular link chain found in block %u of index \"%s\"",
current, RelationGetRelationName(state->rel))));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
leftcurrent = current;
current = opaque->btpo_next;
if (state->lowkey)
{
Assert(state->readonly);
pfree(state->lowkey);
state->lowkey = NULL;
}
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* Copy current target high key as the low key of right sibling.
* Allocate memory in upper level context, so it would be cleared
* after reset of target context.
*
* We only need the low key in corner cases of checking child high
* keys. We use high key only when incomplete split on the child level
* falls to the boundary of pages on the target level. See
* bt_child_highkey_check() for details. So, typically we won't end
* up doing anything with low key, but it's simpler for general case
* high key verification to always have it available.
*
* The correctness of managing low key in the case of concurrent
* splits wasn't investigated yet. Thankfully we only need low key
* for readonly verification and concurrent splits won't happen.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*/
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (state->readonly && !P_RIGHTMOST(opaque))
{
IndexTuple itup;
ItemId itemid;
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, P_HIKEY);
itup = (IndexTuple) PageGetItem(state->target, itemid);
state->lowkey = MemoryContextAlloc(oldcontext, IndexTupleSize(itup));
memcpy(state->lowkey, itup, IndexTupleSize(itup));
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Free page and associated memory for this iteration */
MemoryContextReset(state->targetcontext);
}
while (current != P_NONE);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (state->lowkey)
{
Assert(state->readonly);
pfree(state->lowkey);
state->lowkey = NULL;
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Don't change context for caller */
MemoryContextSwitchTo(oldcontext);
return nextleveldown;
}
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
/*
* Raise an error when target page's left link does not point back to the
* previous target page, called leftcurrent here. The leftcurrent page's
* right link was followed to get to the current target page, and we expect
* mutual agreement among leftcurrent and the current target page. Make sure
* that this condition has definitely been violated in the !readonly case,
* where concurrent page splits are something that we need to deal with.
*
* Cross-page inconsistencies involving pages that don't agree about being
* siblings are known to be a particularly good indicator of corruption
* involving partial writes/lost updates. The bt_right_page_check_scankey
* check also provides a way of detecting cross-page inconsistencies for
* !readonly callers, but it can only detect sibling pages that have an
* out-of-order keyspace, which can't catch many of the problems that we
* expect to catch here.
*
* The classic example of the kind of inconsistency that we can only catch
* with this check (when in !readonly mode) involves three sibling pages that
* were affected by a faulty page split at some point in the past. The
* effects of the split are reflected in the original page and its new right
* sibling page, with a lack of any accompanying changes for the _original_
* right sibling page. The original right sibling page's left link fails to
* point to the new right sibling page (its left link still points to the
* original page), even though the first phase of a page split is supposed to
* work as a single atomic action. This subtle inconsistency will probably
* only break backwards scans in practice.
*
* Note that this is the only place where amcheck will "couple" buffer locks
* (and only for !readonly callers). In general we prefer to avoid more
* thorough cross-page checks in !readonly mode, but it seems worth the
* complexity here. Also, the performance overhead of performing lock
* coupling here is negligible in practice. Control only reaches here with a
* non-corrupt index when there is a concurrent page split at the instant
* caller crossed over to target page from leftcurrent page.
*/
static void
bt_recheck_sibling_links(BtreeCheckState *state,
BlockNumber btpo_prev_from_target,
BlockNumber leftcurrent)
{
if (!state->readonly)
{
Buffer lbuf;
Buffer newtargetbuf;
Page page;
BTPageOpaque opaque;
BlockNumber newtargetblock;
/* Couple locks in the usual order for nbtree: Left to right */
lbuf = ReadBufferExtended(state->rel, MAIN_FORKNUM, leftcurrent,
RBM_NORMAL, state->checkstrategy);
LockBuffer(lbuf, BT_READ);
_bt_checkpage(state->rel, lbuf);
page = BufferGetPage(lbuf);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(page);
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
if (P_ISDELETED(opaque))
{
/*
* Cannot reason about concurrently deleted page -- the left link
* in the page to the right is expected to point to some other
* page to the left (not leftcurrent page).
*
* Note that we deliberately don't give up with a half-dead page.
*/
UnlockReleaseBuffer(lbuf);
return;
}
newtargetblock = opaque->btpo_next;
/* Avoid self-deadlock when newtargetblock == leftcurrent */
if (newtargetblock != leftcurrent)
{
newtargetbuf = ReadBufferExtended(state->rel, MAIN_FORKNUM,
newtargetblock, RBM_NORMAL,
state->checkstrategy);
LockBuffer(newtargetbuf, BT_READ);
_bt_checkpage(state->rel, newtargetbuf);
page = BufferGetPage(newtargetbuf);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(page);
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
/* btpo_prev_from_target may have changed; update it */
btpo_prev_from_target = opaque->btpo_prev;
}
else
{
/*
* leftcurrent right sibling points back to leftcurrent block.
* Index is corrupt. Easiest way to handle this is to pretend
* that we actually read from a distinct page that has an invalid
* block number in its btpo_prev.
*/
newtargetbuf = InvalidBuffer;
btpo_prev_from_target = InvalidBlockNumber;
}
/*
* No need to check P_ISDELETED here, since new target block cannot be
* marked deleted as long as we hold a lock on lbuf
*/
if (BufferIsValid(newtargetbuf))
UnlockReleaseBuffer(newtargetbuf);
UnlockReleaseBuffer(lbuf);
if (btpo_prev_from_target == leftcurrent)
{
/* Report split in left sibling, not target (or new target) */
ereport(DEBUG1,
(errcode(ERRCODE_INTERNAL_ERROR),
Use errmsg_internal for debug messages An inconsistent set of debug-level messages was not using errmsg_internal(), thus uselessly exposing the messages to translation work. Fix those.
2021-02-17 11:24:46 +01:00
errmsg_internal("harmless concurrent page split detected in index \"%s\"",
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u new right sibling=%u original right sibling=%u.",
leftcurrent, newtargetblock,
state->targetblock)));
return;
}
/*
* Index is corrupt. Make sure that we report correct target page.
*
* This could have changed in cases where there was a concurrent page
* split, as well as index corruption (at least in theory). Note that
* btpo_prev_from_target was already updated above.
*/
state->targetblock = newtargetblock;
}
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("left link/right link pair in index \"%s\" not in agreement",
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u left block=%u left link from block=%u.",
state->targetblock, leftcurrent,
btpo_prev_from_target)));
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Function performs the following checks on target page, or pages ancillary to
* target page:
*
* - That every "real" data item is less than or equal to the high key, which
Correct obsolete amcheck comments. Oversight in commit dd299df8.
2019-06-24 19:31:58 +02:00
* is an upper bound on the items on the page. Data items should be
* strictly less than the high key when the page is an internal page.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Correct obsolete amcheck comments. Oversight in commit dd299df8.
2019-06-24 19:31:58 +02:00
* - That within the page, every data item is strictly less than the item
* immediately to its right, if any (i.e., that the items are in order
* within the page, so that the binary searches performed by index scans are
* sane).
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Correct obsolete amcheck comments. Oversight in commit dd299df8.
2019-06-24 19:31:58 +02:00
* - That the last data item stored on the page is strictly less than the
* first data item on the page to the right (when such a first item is
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* available).
*
Correct obsolete amcheck comments. Oversight in commit dd299df8.
2019-06-24 19:31:58 +02:00
* - Various checks on the structure of tuples themselves. For example, check
* that non-pivot tuples have no truncated attributes.
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
*
Add amcheck missing downlink tests. Also use palloc0() for main amcheck state, and adjust a few comments. Somehow I pushed old version of patch in commit 4eaf7eaccb291f5d32d9f05284bfca9c5744de9a, so commit the difference. Peter Geoghegan
2018-04-25 21:05:53 +02:00
* Furthermore, when state passed shows ShareLock held, function also checks:
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Correct obsolete amcheck comments. Oversight in commit dd299df8.
2019-06-24 19:31:58 +02:00
* - That all child pages respect strict lower bound from parent's pivot
* tuple.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Add amcheck missing downlink tests. Also use palloc0() for main amcheck state, and adjust a few comments. Somehow I pushed old version of patch in commit 4eaf7eaccb291f5d32d9f05284bfca9c5744de9a, so commit the difference. Peter Geoghegan
2018-04-25 21:05:53 +02:00
* - That downlink to block was encountered in parent where that's expected.
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
*
* - That high keys of child pages matches corresponding pivot keys in parent.
Add amcheck missing downlink tests. Also use palloc0() for main amcheck state, and adjust a few comments. Somehow I pushed old version of patch in commit 4eaf7eaccb291f5d32d9f05284bfca9c5744de9a, so commit the difference. Peter Geoghegan
2018-04-25 21:05:53 +02:00
*
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* This is also where heapallindexed callers use their Bloom filter to
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* fingerprint IndexTuples for later table_index_build_scan() verification.
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
*
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* Note: Memory allocated in this routine is expected to be released by caller
* resetting state->targetcontext.
*/
static void
bt_target_page_check(BtreeCheckState *state)
{
OffsetNumber offset;
OffsetNumber max;
BTPageOpaque topaque;
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
topaque = BTPageGetOpaque(state->target);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
max = PageGetMaxOffsetNumber(state->target);
elog(DEBUG2, "verifying %u items on %s block %u", max,
P_ISLEAF(topaque) ? "leaf" : "internal", state->targetblock);
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
/*
* Check the number of attributes in high key. Note, rightmost page
* doesn't contain a high key, so nothing to check
*/
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
if (!P_RIGHTMOST(topaque))
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
{
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
ItemId itemid;
IndexTuple itup;
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
/* Verify line pointer before checking tuple */
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, P_HIKEY);
if (!_bt_check_natts(state->rel, state->heapkeyspace, state->target,
P_HIKEY))
{
itup = (IndexTuple) PageGetItem(state->target, itemid);
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("wrong number of high key index tuple attributes in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index block=%u natts=%u block type=%s page lsn=%X/%X.",
state->targetblock,
BTreeTupleGetNAtts(itup, state->rel),
P_ISLEAF(topaque) ? "heap" : "index",
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
LSN_FORMAT_ARGS(state->targetlsn))));
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
}
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Loop over page items, starting from first non-highkey item, not high
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
* key (if any). Most tests are not performed for the "negative infinity"
* real item (if any).
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
for (offset = P_FIRSTDATAKEY(topaque);
offset <= max;
offset = OffsetNumberNext(offset))
{
ItemId itemid;
IndexTuple itup;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
size_t tupsize;
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
BTScanInsert skey;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
bool lowersizelimit;
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer scantid;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
CHECK_FOR_INTERRUPTS();
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, offset);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
itup = (IndexTuple) PageGetItem(state->target, itemid);
tupsize = IndexTupleSize(itup);
/*
* lp_len should match the IndexTuple reported length exactly, since
* lp_len is completely redundant in indexes, and both sources of
* tuple length are MAXALIGN()'d. nbtree does not use lp_len all that
* frequently, and is surprisingly tolerant of corrupt lp_len fields.
*/
if (tupsize != ItemIdGetLength(itemid))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index tuple size does not equal lp_len in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=(%u,%u) tuple size=%zu lp_len=%u page lsn=%X/%X.",
state->targetblock, offset,
tupsize, ItemIdGetLength(itemid),
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
LSN_FORMAT_ARGS(state->targetlsn)),
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
errhint("This could be a torn page problem.")));
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
/* Check the number of index tuple attributes */
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
if (!_bt_check_natts(state->rel, state->heapkeyspace, state->target,
offset))
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer tid;
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
char *itid,
*htid;
itid = psprintf("(%u,%u)", state->targetblock, offset);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
tid = BTreeTupleGetPointsToTID(itup);
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
htid = psprintf("(%u,%u)",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointerGetBlockNumberNoCheck(tid),
ItemPointerGetOffsetNumberNoCheck(tid));
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
errmsg("wrong number of index tuple attributes in index \"%s\"",
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=%s natts=%u points to %s tid=%s page lsn=%X/%X.",
itid,
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
BTreeTupleGetNAtts(itup, state->rel),
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
P_ISLEAF(topaque) ? "heap" : "index",
htid,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
* Don't try to generate scankey using "negative infinity" item on
* internal pages. They are always truncated to zero attributes.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
if (offset_is_negative_infinity(topaque, offset))
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
{
/*
* We don't call bt_child_check() for "negative infinity" items.
* But if we're performing downlink connectivity check, we do it
* for every item including "negative infinity" one.
*/
if (!P_ISLEAF(topaque) && state->readonly)
{
bt_child_highkey_check(state,
offset,
NULL,
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
topaque->btpo_level);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
continue;
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
/*
* Readonly callers may optionally verify that non-pivot tuples can
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
* each be found by an independent search that starts from the root.
* Note that we deliberately don't do individual searches for each
* TID, since the posting list itself is validated by other checks.
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
*/
if (state->rootdescend && P_ISLEAF(topaque) &&
!bt_rootdescend(state, itup))
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer tid = BTreeTupleGetPointsToTID(itup);
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
char *itid,
*htid;
itid = psprintf("(%u,%u)", state->targetblock, offset);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
htid = psprintf("(%u,%u)", ItemPointerGetBlockNumber(tid),
ItemPointerGetOffsetNumber(tid));
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("could not find tuple using search from root page in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=%s points to heap tid=%s page lsn=%X/%X.",
itid, htid,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
}
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/*
* If tuple is a posting list tuple, make sure posting list TIDs are
* in order
*/
if (BTreeTupleIsPosting(itup))
{
ItemPointerData last;
ItemPointer current;
ItemPointerCopy(BTreeTupleGetHeapTID(itup), &last);
for (int i = 1; i < BTreeTupleGetNPosting(itup); i++)
{
current = BTreeTupleGetPostingN(itup, i);
if (ItemPointerCompare(current, &last) <= 0)
{
char *itid = psprintf("(%u,%u)", state->targetblock, offset);
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("posting list contains misplaced TID in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=%s posting list offset=%d page lsn=%X/%X.",
itid, i,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
}
ItemPointerCopy(current, &last);
}
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Build insertion scankey for current page offset */
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
skey = bt_mkscankey_pivotsearch(state->rel, state->heaprel, itup);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/*
* Make sure tuple size does not exceed the relevant BTREE_VERSION
* specific limit.
*
* BTREE_VERSION 4 (which introduced heapkeyspace rules) requisitioned
* a small amount of space from BTMaxItemSize() in order to ensure
* that suffix truncation always has enough space to add an explicit
* heap TID back to a tuple -- we pessimistically assume that every
* newly inserted tuple will eventually need to have a heap TID
* appended during a future leaf page split, when the tuple becomes
* the basis of the new high key (pivot tuple) for the leaf page.
*
* Since the reclaimed space is reserved for that purpose, we must not
* enforce the slightly lower limit when the extra space has been used
* as intended. In other words, there is only a cross-version
* difference in the limit on tuple size within leaf pages.
*
* Still, we're particular about the details within BTREE_VERSION 4
* internal pages. Pivot tuples may only use the extra space for its
* designated purpose. Enforce the lower limit for pivot tuples when
* an explicit heap TID isn't actually present. (In all other cases
* suffix truncation is guaranteed to generate a pivot tuple that's no
Harmonize nbtree page split point code. An nbtree split point can be thought of as a point between two adjoining tuples from an imaginary version of the page being split that includes the incoming/new item (in addition to the items that really are on the page). These adjoining tuples are called the lastleft and firstright tuples. The variables that represent split points contained a field called firstright, which is an offset number of the first data item from the original page that goes on the new right page. The corresponding tuple from origpage was usually the same thing as the actual firstright tuple, but not always: the firstright tuple is sometimes the new/incoming item instead. This situation seems unnecessarily confusing. Make things clearer by renaming the origpage offset returned by _bt_findsplitloc() to "firstrightoff". We now have a firstright tuple and a firstrightoff offset number which are comparable to the newitem/lastleft tuples and the newitemoff/lastleftoff offset numbers respectively. Also make sure that we are consistent about how we describe nbtree page split point state. Push the responsibility for dealing with pg_upgrade'd !heapkeyspace indexes down to lower level code, relieving _bt_split() from dealing with it directly. This means that we always have a palloc'd left page high key on the leaf level, no matter what. This enables simplifying some of the code (and code comments) within _bt_split(). Finally, restructure the page split code to make it clearer why suffix truncation (which only takes place during leaf page splits) is completely different to the first data item truncation that takes place during internal page splits. Tuples are marked as having fewer attributes stored in both cases, and the firstright tuple is truncated in both cases, so it's easy to imagine somebody missing the distinction.
2020-04-14 01:39:55 +02:00
* larger than the firstright tuple provided to it by its caller.)
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
*/
lowersizelimit = skey->heapkeyspace &&
(P_ISLEAF(topaque) || BTreeTupleGetHeapTID(itup) == NULL);
if (tupsize > (lowersizelimit ? BTMaxItemSize(state->target) :
BTMaxItemSizeNoHeapTid(state->target)))
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer tid = BTreeTupleGetPointsToTID(itup);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
char *itid,
*htid;
itid = psprintf("(%u,%u)", state->targetblock, offset);
htid = psprintf("(%u,%u)",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointerGetBlockNumberNoCheck(tid),
ItemPointerGetOffsetNumberNoCheck(tid));
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index row size %zu exceeds maximum for index \"%s\"",
tupsize, RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=%s points to %s tid=%s page lsn=%X/%X.",
itid,
P_ISLEAF(topaque) ? "heap" : "index",
htid,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* Fingerprint leaf page tuples (those that point to the heap) */
if (state->heapallindexed && P_ISLEAF(topaque) && !ItemIdIsDead(itemid))
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
{
IndexTuple norm;
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
if (BTreeTupleIsPosting(itup))
{
/* Fingerprint all elements as distinct "plain" tuples */
for (int i = 0; i < BTreeTupleGetNPosting(itup); i++)
{
IndexTuple logtuple;
logtuple = bt_posting_plain_tuple(itup, i);
norm = bt_normalize_tuple(state, logtuple);
bloom_add_element(state->filter, (unsigned char *) norm,
IndexTupleSize(norm));
/* Be tidy */
if (norm != logtuple)
pfree(norm);
pfree(logtuple);
}
}
else
{
norm = bt_normalize_tuple(state, itup);
bloom_add_element(state->filter, (unsigned char *) norm,
IndexTupleSize(norm));
/* Be tidy */
if (norm != itup)
pfree(norm);
}
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
}
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* * High key check *
*
* If there is a high key (if this is not the rightmost page on its
* entire level), check that high key actually is upper bound on all
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
* page items. If this is a posting list tuple, we'll need to set
* scantid to be highest TID in posting list.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* We prefer to check all items against high key rather than checking
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* just the last and trusting that the operator class obeys the
* transitive law (which implies that all previous items also
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* respected the high key invariant if they pass the item order
* check).
*
* Ideally, we'd compare every item in the index against every other
* item in the index, and not trust opclass obedience of the
* transitive law to bridge the gap between children and their
* grandparents (as well as great-grandparents, and so on). We don't
* go to those lengths because that would be prohibitively expensive,
* and probably not markedly more effective in practice.
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
*
* On the leaf level, we check that the key is <= the highkey.
* However, on non-leaf levels we check that the key is < the highkey,
* because the high key is "just another separator" rather than a copy
* of some existing key item; we expect it to be unique among all keys
* on the same level. (Suffix truncation will sometimes produce a
* leaf highkey that is an untruncated copy of the lastleft item, but
* never any other item, which necessitates weakening the leaf level
* check to <=.)
*
* Full explanation for why a highkey is never truly a copy of another
* item from the same level on internal levels:
*
* While the new left page's high key is copied from the first offset
* on the right page during an internal page split, that's not the
* full story. In effect, internal pages are split in the middle of
* the firstright tuple, not between the would-be lastleft and
* firstright tuples: the firstright key ends up on the left side as
* left's new highkey, and the firstright downlink ends up on the
* right side as right's new "negative infinity" item. The negative
* infinity tuple is truncated to zero attributes, so we're only left
* with the downlink. In other words, the copying is just an
* implementation detail of splitting in the middle of a (pivot)
* tuple. (See also: "Notes About Data Representation" in the nbtree
* README.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
scantid = skey->scantid;
if (state->heapkeyspace && BTreeTupleIsPosting(itup))
skey->scantid = BTreeTupleGetMaxHeapTID(itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (!P_RIGHTMOST(topaque) &&
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
!(P_ISLEAF(topaque) ? invariant_leq_offset(state, skey, P_HIKEY) :
invariant_l_offset(state, skey, P_HIKEY)))
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer tid = BTreeTupleGetPointsToTID(itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
char *itid,
*htid;
itid = psprintf("(%u,%u)", state->targetblock, offset);
htid = psprintf("(%u,%u)",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointerGetBlockNumberNoCheck(tid),
ItemPointerGetOffsetNumberNoCheck(tid));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("high key invariant violated for index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=%s points to %s tid=%s page lsn=%X/%X.",
itid,
P_ISLEAF(topaque) ? "heap" : "index",
htid,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/* Reset, in case scantid was set to (itup) posting tuple's max TID */
skey->scantid = scantid;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* * Item order check *
*
* Check that items are stored on page in logical order, by checking
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* current item is strictly less than next item (if any).
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
if (OffsetNumberNext(offset) <= max &&
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
!invariant_l_offset(state, skey, OffsetNumberNext(offset)))
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer tid;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
char *itid,
*htid,
*nitid,
*nhtid;
itid = psprintf("(%u,%u)", state->targetblock, offset);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
tid = BTreeTupleGetPointsToTID(itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
htid = psprintf("(%u,%u)",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointerGetBlockNumberNoCheck(tid),
ItemPointerGetOffsetNumberNoCheck(tid));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
nitid = psprintf("(%u,%u)", state->targetblock,
OffsetNumberNext(offset));
/* Reuse itup to get pointed-to heap location of second item */
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target,
OffsetNumberNext(offset));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
itup = (IndexTuple) PageGetItem(state->target, itemid);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
tid = BTreeTupleGetPointsToTID(itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
nhtid = psprintf("(%u,%u)",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointerGetBlockNumberNoCheck(tid),
ItemPointerGetOffsetNumberNoCheck(tid));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("item order invariant violated for index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Lower index tid=%s (points to %s tid=%s) "
"higher index tid=%s (points to %s tid=%s) "
"page lsn=%X/%X.",
itid,
P_ISLEAF(topaque) ? "heap" : "index",
htid,
nitid,
P_ISLEAF(topaque) ? "heap" : "index",
nhtid,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
* * Last item check *
*
* Check last item against next/right page's first data item's when
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* last item on page is reached. This additional check will detect
* transposed pages iff the supposed right sibling page happens to
* belong before target in the key space. (Otherwise, a subsequent
* heap verification will probably detect the problem.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* This check is similar to the item order check that will have
* already been performed for every other "real" item on target page
* when last item is checked. The difference is that the next item
* (the item that is compared to target's last item) needs to come
* from the next/sibling page. There may not be such an item
* available from sibling for various reasons, though (e.g., target is
* the rightmost page on level).
*/
else if (offset == max)
{
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
BTScanInsert rightkey;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Get item in next/right page */
rightkey = bt_right_page_check_scankey(state);
if (rightkey &&
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
!invariant_g_offset(state, rightkey, max))
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
/*
* As explained at length in bt_right_page_check_scankey(),
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* there is a known !readonly race that could account for
* apparent violation of invariant, which we must check for
* before actually proceeding with raising error. Our canary
* condition is that target page was deleted.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
if (!state->readonly)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
/* Get fresh copy of target page */
state->target = palloc_btree_page(state, state->targetblock);
/* Note that we deliberately do not update target LSN */
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
topaque = BTPageGetOpaque(state->target);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* All !readonly checks now performed; just return
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
if (P_IGNORE(topaque))
return;
}
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("cross page item order invariant violated for index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Last item on page tid=(%u,%u) page lsn=%X/%X.",
state->targetblock, offset,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
}
/*
* * Downlink check *
*
* Additional check of child items iff this is an internal page and
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* caller holds a ShareLock. This happens for every downlink (item)
* in target excluding the negative-infinity downlink (again, this is
* because it has no useful value to compare).
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
if (!P_ISLEAF(topaque) && state->readonly)
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
bt_child_check(state, skey, offset);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* Special case bt_child_highkey_check() call
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*
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
* We don't pass a real downlink, but we've to finish the level
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* processing. If condition is satisfied, we've already processed all the
* downlinks from the target level. But there still might be pages to the
* right of the child page pointer to by our rightmost downlink. And they
* might have missing downlinks. This final call checks for them.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*/
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (!P_ISLEAF(topaque) && P_RIGHTMOST(topaque) && state->readonly)
{
bt_child_highkey_check(state, InvalidOffsetNumber,
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
NULL, topaque->btpo_level);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
* Return a scankey for an item on page to right of current target (or the
* first non-ignorable page), sufficient to check ordering invariant on last
* item in current target page. Returned scankey relies on local memory
* allocated for the child page, which caller cannot pfree(). Caller's memory
* context should be reset between calls here.
*
* This is the first data item, and so all adjacent items are checked against
* their immediate sibling item (which may be on a sibling page, or even a
* "cousin" page at parent boundaries where target's rightlink points to page
* with different parent page). If no such valid item is available, return
* NULL instead.
*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* Note that !readonly callers must reverify that target page has not
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* been concurrently deleted.
*/
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
static BTScanInsert
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
bt_right_page_check_scankey(BtreeCheckState *state)
{
BTPageOpaque opaque;
ItemId rightitem;
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
IndexTuple firstitup;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
BlockNumber targetnext;
Page rightpage;
OffsetNumber nline;
/* Determine target's next block number */
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(state->target);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* If target is already rightmost, no right sibling; nothing to do here */
if (P_RIGHTMOST(opaque))
return NULL;
/*
* General notes on concurrent page splits and page deletion:
*
* Routines like _bt_search() don't require *any* page split interlock
* when descending the tree, including something very light like a buffer
* pin. That's why it's okay that we don't either. This avoidance of any
* need to "couple" buffer locks is the raison d' etre of the Lehman & Yao
* algorithm, in fact.
*
* That leaves deletion. A deleted page won't actually be recycled by
* VACUUM early enough for us to fail to at least follow its right link
* (or left link, or downlink) and find its sibling, because recycling
* does not occur until no possible index scan could land on the page.
* Index scans can follow links with nothing more than their snapshot as
* an interlock and be sure of at least that much. (See page
snapshot scalability: Don't compute global horizons while building snapshots. To make GetSnapshotData() more scalable, it cannot not look at at each proc's xmin: While snapshot contents do not need to change whenever a read-only transaction commits or a snapshot is released, a proc's xmin is modified in those cases. The frequency of xmin modifications leads to, particularly on higher core count systems, many cache misses inside GetSnapshotData(), despite the data underlying a snapshot not changing. That is the most significant source of GetSnapshotData() scaling poorly on larger systems. Without accessing xmins, GetSnapshotData() cannot calculate accurate horizons / thresholds as it has so far. But we don't really have to: The horizons don't actually change that much between GetSnapshotData() calls. Nor are the horizons actually used every time a snapshot is built. The trick this commit introduces is to delay computation of accurate horizons until there use and using horizon boundaries to determine whether accurate horizons need to be computed. The use of RecentGlobal[Data]Xmin to decide whether a row version could be removed has been replaces with new GlobalVisTest* functions. These use two thresholds to determine whether a row can be pruned: 1) definitely_needed, indicating that rows deleted by XIDs >= definitely_needed are definitely still visible. 2) maybe_needed, indicating that rows deleted by XIDs < maybe_needed can definitely be removed GetSnapshotData() updates definitely_needed to be the xmin of the computed snapshot. When testing whether a row can be removed (with GlobalVisTestIsRemovableXid()) and the tested XID falls in between the two (i.e. XID >= maybe_needed && XID < definitely_needed) the boundaries can be recomputed to be more accurate. As it is not cheap to compute accurate boundaries, we limit the number of times that happens in short succession. As the boundaries used by GlobalVisTestIsRemovableXid() are never reset (with maybe_needed updated by GetSnapshotData()), it is likely that further test can benefit from an earlier computation of accurate horizons. To avoid regressing performance when old_snapshot_threshold is set (as that requires an accurate horizon to be computed), heap_page_prune_opt() doesn't unconditionally call TransactionIdLimitedForOldSnapshots() anymore. Both the computation of the limited horizon, and the triggering of errors (with SetOldSnapshotThresholdTimestamp()) is now only done when necessary to remove tuples. This commit just removes the accesses to PGXACT->xmin from GetSnapshotData(), but other members of PGXACT residing in the same cache line are accessed. Therefore this in itself does not result in a significant improvement. Subsequent commits will take advantage of the fact that GetSnapshotData() now does not need to access xmins anymore. Note: This contains a workaround in heap_page_prune_opt() to keep the snapshot_too_old tests working. While that workaround is ugly, the tests currently are not meaningful, and it seems best to address them separately. Author: Andres Freund <andres@anarazel.de> Reviewed-By: Robert Haas <robertmhaas@gmail.com> Reviewed-By: Thomas Munro <thomas.munro@gmail.com> Reviewed-By: David Rowley <dgrowleyml@gmail.com> Discussion: https://postgr.es/m/20200301083601.ews6hz5dduc3w2se@alap3.anarazel.de
2020-08-13 01:03:49 +02:00
* recycling/"visible to everyone" notes in nbtree README.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Furthermore, it's okay if we follow a rightlink and find a half-dead or
* dead (ignorable) page one or more times. There will either be a
* further right link to follow that leads to a live page before too long
* (before passing by parent's rightmost child), or we will find the end
* of the entire level instead (possible when parent page is itself the
* rightmost on its level).
*/
targetnext = opaque->btpo_next;
for (;;)
{
CHECK_FOR_INTERRUPTS();
rightpage = palloc_btree_page(state, targetnext);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(rightpage);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (!P_IGNORE(opaque) || P_RIGHTMOST(opaque))
break;
amcheck: Reduce debug message verbosity. Empty sibling pages can occasionally be much more common than any other event that we report on at elevel DEBUG1. Increase the elevel for relevant cases to DEBUG2 to avoid overwhelming the user with relatively insignificant details.
2021-03-16 21:11:17 +01:00
/*
* We landed on a deleted or half-dead sibling page. Step right until
* we locate a live sibling page.
*/
ereport(DEBUG2,
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
(errcode(ERRCODE_NO_DATA),
amcheck: Reduce debug message verbosity. Empty sibling pages can occasionally be much more common than any other event that we report on at elevel DEBUG1. Increase the elevel for relevant cases to DEBUG2 to avoid overwhelming the user with relatively insignificant details.
2021-03-16 21:11:17 +01:00
errmsg_internal("level %u sibling page in block %u of index \"%s\" was found deleted or half dead",
opaque->btpo_level, targetnext, RelationGetRelationName(state->rel)),
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
errdetail_internal("Deleted page found when building scankey from right sibling.")));
amcheck: Reduce debug message verbosity. Empty sibling pages can occasionally be much more common than any other event that we report on at elevel DEBUG1. Increase the elevel for relevant cases to DEBUG2 to avoid overwhelming the user with relatively insignificant details.
2021-03-16 21:11:17 +01:00
targetnext = opaque->btpo_next;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/* Be slightly more pro-active in freeing this memory, just in case */
pfree(rightpage);
}
/*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* No ShareLock held case -- why it's safe to proceed.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Problem:
*
* We must avoid false positive reports of corruption when caller treats
* item returned here as an upper bound on target's last item. In
* general, false positives are disallowed. Avoiding them here when
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* caller is !readonly is subtle.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* A concurrent page deletion by VACUUM of the target page can result in
* the insertion of items on to this right sibling page that would
* previously have been inserted on our target page. There might have
* been insertions that followed the target's downlink after it was made
* to point to right sibling instead of target by page deletion's first
* phase. The inserters insert items that would belong on target page.
* This race is very tight, but it's possible. This is our only problem.
*
* Non-problems:
*
* We are not hindered by a concurrent page split of the target; we'll
* never land on the second half of the page anyway. A concurrent split
* of the right page will also not matter, because the first data item
* remains the same within the left half, which we'll reliably land on. If
* we had to skip over ignorable/deleted pages, it cannot matter because
* their key space has already been atomically merged with the first
* non-ignorable page we eventually find (doesn't matter whether the page
* we eventually find is a true sibling or a cousin of target, which we go
* into below).
*
* Solution:
*
* Caller knows that it should reverify that target is not ignorable
* (half-dead or deleted) when cross-page sibling item comparison appears
* to indicate corruption (invariant fails). This detects the single race
* condition that exists for caller. This is correct because the
* continued existence of target block as non-ignorable (not half-dead or
* deleted) implies that target page was not merged into from the right by
* deletion; the key space at or after target never moved left. Target's
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* parent either has the same downlink to target as before, or a <
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* downlink due to deletion at the left of target. Target either has the
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* same highkey as before, or a highkey < before when there is a page
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* split. (The rightmost concurrently-split-from-target-page page will
* still have the same highkey as target was originally found to have,
* which for our purposes is equivalent to target's highkey itself never
* changing, since we reliably skip over
* concurrently-split-from-target-page pages.)
*
* In simpler terms, we allow that the key space of the target may expand
* left (the key space can move left on the left side of target only), but
* the target key space cannot expand right and get ahead of us without
* our detecting it. The key space of the target cannot shrink, unless it
* shrinks to zero due to the deletion of the original page, our canary
* condition. (To be very precise, we're a bit stricter than that because
* it might just have been that the target page split and only the
* original target page was deleted. We can be more strict, just not more
* lax.)
*
* Top level tree walk caller moves on to next page (makes it the new
* target) following recovery from this race. (cf. The rationale for
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* child/downlink verification needing a ShareLock within
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* bt_child_check(), where page deletion is also the main source of
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* trouble.)
*
* Note that it doesn't matter if right sibling page here is actually a
* cousin page, because in order for the key space to be readjusted in a
* way that causes us issues in next level up (guiding problematic
* concurrent insertions to the cousin from the grandparent rather than to
* the sibling from the parent), there'd have to be page deletion of
* target's parent page (affecting target's parent's downlink in target's
* grandparent page). Internal page deletion only occurs when there are
* no child pages (they were all fully deleted), and caller is checking
* that the target's parent has at least one non-deleted (so
* non-ignorable) child: the target page. (Note that the first phase of
* deletion atomically marks the page to be deleted half-dead/ignorable at
* the same time downlink in its parent is removed, so caller will
* definitely not fail to detect that this happened.)
*
* This trick is inspired by the method backward scans use for dealing
* with concurrent page splits; concurrent page deletion is a problem that
* similarly receives special consideration sometimes (it's possible that
* the backwards scan will re-read its "original" block after failing to
* find a right-link to it, having already moved in the opposite direction
* (right/"forwards") a few times to try to locate one). Just like us,
* that happens only to determine if there was a concurrent page deletion
* of a reference page, and just like us if there was a page deletion of
* that reference page it means we can move on from caring about the
* reference page. See the nbtree README for a full description of how
* that works.
*/
nline = PageGetMaxOffsetNumber(rightpage);
/*
* Get first data item, if any
*/
if (P_ISLEAF(opaque) && nline >= P_FIRSTDATAKEY(opaque))
{
/* Return first data item (if any) */
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
rightitem = PageGetItemIdCareful(state, targetnext, rightpage,
P_FIRSTDATAKEY(opaque));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
else if (!P_ISLEAF(opaque) &&
nline >= OffsetNumberNext(P_FIRSTDATAKEY(opaque)))
{
/*
* Return first item after the internal page's "negative infinity"
* item
*/
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
rightitem = PageGetItemIdCareful(state, targetnext, rightpage,
OffsetNumberNext(P_FIRSTDATAKEY(opaque)));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
else
{
/*
* No first item. Page is probably empty leaf page, but it's also
* possible that it's an internal page with only a negative infinity
* item.
*/
amcheck: Reduce debug message verbosity. Empty sibling pages can occasionally be much more common than any other event that we report on at elevel DEBUG1. Increase the elevel for relevant cases to DEBUG2 to avoid overwhelming the user with relatively insignificant details.
2021-03-16 21:11:17 +01:00
ereport(DEBUG2,
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
(errcode(ERRCODE_NO_DATA),
Use errmsg_internal for debug messages An inconsistent set of debug-level messages was not using errmsg_internal(), thus uselessly exposing the messages to translation work. Fix those.
2021-02-17 11:24:46 +01:00
errmsg_internal("%s block %u of index \"%s\" has no first data item",
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
P_ISLEAF(opaque) ? "leaf" : "internal", targetnext,
RelationGetRelationName(state->rel))));
return NULL;
}
/*
* Return first real item scankey. Note that this relies on right page
* memory remaining allocated.
*/
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
firstitup = (IndexTuple) PageGetItem(rightpage, rightitem);
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
return bt_mkscankey_pivotsearch(state->rel, state->heaprel, firstitup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/*
* Check if two tuples are binary identical except the block number. So,
* this function is capable to compare pivot keys on different levels.
*/
static bool
Fix amcheck child check pg_upgrade bug. Commit d114cc53 overlooked the fact that pg_upgrade'd B-Tree indexes have leaf page high keys whose offset numbers do not match the one from the copy of the tuple one level up (the copy stored with a downlink for leaf page's right sibling page). This led to false positive reports of corruption from bt_index_parent_check() when it was called to verify a pg_upgrade'd index. To fix, skip comparing the offset number on pg_upgrade'd B-Tree indexes. Author: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Author: Peter Geoghegan <pg@bowt.ie> Reported-By: Andrew Bille <andrewbille@gmail.com> Diagnosed-By: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Bug: #16619 Discussion: https://postgr.es/m/16619-aaba10f83fdc1c3c@postgresql.org Backpatch: 13-, where child check was enhanced.
2020-09-16 19:42:30 +02:00
bt_pivot_tuple_identical(bool heapkeyspace, IndexTuple itup1, IndexTuple itup2)
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
{
if (IndexTupleSize(itup1) != IndexTupleSize(itup2))
return false;
Fix amcheck child check pg_upgrade bug. Commit d114cc53 overlooked the fact that pg_upgrade'd B-Tree indexes have leaf page high keys whose offset numbers do not match the one from the copy of the tuple one level up (the copy stored with a downlink for leaf page's right sibling page). This led to false positive reports of corruption from bt_index_parent_check() when it was called to verify a pg_upgrade'd index. To fix, skip comparing the offset number on pg_upgrade'd B-Tree indexes. Author: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Author: Peter Geoghegan <pg@bowt.ie> Reported-By: Andrew Bille <andrewbille@gmail.com> Diagnosed-By: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Bug: #16619 Discussion: https://postgr.es/m/16619-aaba10f83fdc1c3c@postgresql.org Backpatch: 13-, where child check was enhanced.
2020-09-16 19:42:30 +02:00
if (heapkeyspace)
{
/*
* Offset number will contain important information in heapkeyspace
* indexes: the number of attributes left in the pivot tuple following
* suffix truncation. Don't skip over it (compare it too).
*/
if (memcmp(&itup1->t_tid.ip_posid, &itup2->t_tid.ip_posid,
IndexTupleSize(itup1) -
offsetof(ItemPointerData, ip_posid)) != 0)
return false;
}
else
{
/*
* Cannot rely on offset number field having consistent value across
* levels on pg_upgrade'd !heapkeyspace indexes. Compare contents of
* tuple starting from just after item pointer (i.e. after block
* number and offset number).
*/
if (memcmp(&itup1->t_info, &itup2->t_info,
IndexTupleSize(itup1) -
offsetof(IndexTupleData, t_info)) != 0)
return false;
}
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
return true;
}
/*---
* Check high keys on the child level. Traverse rightlinks from previous
* downlink to the current one. Check that there are no intermediate pages
* with missing downlinks.
*
* If 'loaded_child' is given, it's assumed to be the page pointed to by the
* downlink referenced by 'downlinkoffnum' of the target page.
*
* Basically this function is called for each target downlink and checks two
* invariants:
*
* 1) You can reach the next child from previous one via rightlinks;
* 2) Each child high key have matching pivot key on target level.
*
* Consider the sample tree picture.
*
* 1
* / \
* 2 <-> 3
* / \ / \
* 4 <> 5 <> 6 <> 7 <> 8
*
* This function will be called for blocks 4, 5, 6 and 8. Consider what is
* happening for each function call.
*
* - The function call for block 4 initializes data structure and matches high
* key of block 4 to downlink's pivot key of block 2.
* - The high key of block 5 is matched to the high key of block 2.
* - The block 6 has an incomplete split flag set, so its high key isn't
* matched to anything.
* - The function call for block 8 checks that block 8 can be found while
* following rightlinks from block 6. The high key of block 7 will be
* matched to downlink's pivot key in block 3.
*
* There is also final call of this function, which checks that there is no
* missing downlinks for children to the right of the child referenced by
* rightmost downlink in target level.
*/
static void
bt_child_highkey_check(BtreeCheckState *state,
OffsetNumber target_downlinkoffnum,
Page loaded_child,
uint32 target_level)
{
BlockNumber blkno = state->prevrightlink;
Page page;
BTPageOpaque opaque;
bool rightsplit = state->previncompletesplit;
bool first = true;
ItemId itemid;
IndexTuple itup;
BlockNumber downlink;
if (OffsetNumberIsValid(target_downlinkoffnum))
{
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, target_downlinkoffnum);
itup = (IndexTuple) PageGetItem(state->target, itemid);
downlink = BTreeTupleGetDownLink(itup);
}
else
{
downlink = P_NONE;
}
/*
* If no previous rightlink is memorized for current level just below
* target page's level, we are about to start from the leftmost page. We
* can't follow rightlinks from previous page, because there is no
* previous page. But we still can match high key.
*
* So we initialize variables for the loop above like there is previous
* page referencing current child. Also we imply previous page to not
* have incomplete split flag, that would make us require downlink for
* current child. That's correct, because leftmost page on the level
* should always have parent downlink.
*/
if (!BlockNumberIsValid(blkno))
{
blkno = downlink;
rightsplit = false;
}
/* Move to the right on the child level */
while (true)
{
/*
* Did we traverse the whole tree level and this is check for pages to
* the right of rightmost downlink?
*/
if (blkno == P_NONE && downlink == P_NONE)
{
state->prevrightlink = InvalidBlockNumber;
state->previncompletesplit = false;
return;
}
/* Did we traverse the whole tree level and don't find next downlink? */
if (blkno == P_NONE)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("can't traverse from downlink %u to downlink %u of index \"%s\"",
state->prevrightlink, downlink,
RelationGetRelationName(state->rel))));
/* Load page contents */
if (blkno == downlink && loaded_child)
page = loaded_child;
else
page = palloc_btree_page(state, blkno);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(page);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/* The first page we visit at the level should be leftmost */
if (first && !BlockNumberIsValid(state->prevrightlink) && !P_LEFTMOST(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("the first child of leftmost target page is not leftmost of its level in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Target block=%u child block=%u target page lsn=%X/%X.",
state->targetblock, blkno,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
/* Do level sanity check */
if ((!P_ISDELETED(opaque) || P_HAS_FULLXID(opaque)) &&
opaque->btpo_level != target_level - 1)
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("block found while following rightlinks from child of index \"%s\" has invalid level",
RelationGetRelationName(state->rel)),
errdetail_internal("Block pointed to=%u expected level=%u level in pointed to block=%u.",
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
blkno, target_level - 1, opaque->btpo_level)));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/* Try to detect circular links */
if ((!first && blkno == state->prevrightlink) || blkno == opaque->btpo_prev)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("circular link chain found in block %u of index \"%s\"",
blkno, RelationGetRelationName(state->rel))));
if (blkno != downlink && !P_IGNORE(opaque))
{
/* blkno probably has missing parent downlink */
bt_downlink_missing_check(state, rightsplit, blkno, page);
}
rightsplit = P_INCOMPLETE_SPLIT(opaque);
/*
Fix amcheck child check pg_upgrade bug. Commit d114cc53 overlooked the fact that pg_upgrade'd B-Tree indexes have leaf page high keys whose offset numbers do not match the one from the copy of the tuple one level up (the copy stored with a downlink for leaf page's right sibling page). This led to false positive reports of corruption from bt_index_parent_check() when it was called to verify a pg_upgrade'd index. To fix, skip comparing the offset number on pg_upgrade'd B-Tree indexes. Author: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Author: Peter Geoghegan <pg@bowt.ie> Reported-By: Andrew Bille <andrewbille@gmail.com> Diagnosed-By: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Bug: #16619 Discussion: https://postgr.es/m/16619-aaba10f83fdc1c3c@postgresql.org Backpatch: 13-, where child check was enhanced.
2020-09-16 19:42:30 +02:00
* If we visit page with high key, check that it is equal to the
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* target key next to corresponding downlink.
*/
if (!rightsplit && !P_RIGHTMOST(opaque))
{
BTPageOpaque topaque;
IndexTuple highkey;
OffsetNumber pivotkey_offset;
/* Get high key */
itemid = PageGetItemIdCareful(state, blkno, page, P_HIKEY);
highkey = (IndexTuple) PageGetItem(page, itemid);
/*
* There might be two situations when we examine high key. If
* current child page is referenced by given target downlink, we
* should look to the next offset number for matching key from
* target page.
*
* Alternatively, we're following rightlinks somewhere in the
* middle between page referenced by previous target's downlink
* and the page referenced by current target's downlink. If
* current child page hasn't incomplete split flag set, then its
* high key should match to the target's key of current offset
* number. This happens when a previous call here (to
* bt_child_highkey_check()) found an incomplete split, and we
* reach a right sibling page without a downlink -- the right
* sibling page's high key still needs to be matched to a
* separator key on the parent/target level.
*
* Don't apply OffsetNumberNext() to target_downlinkoffnum when we
* already had to step right on the child level. Our traversal of
* the child level must try to move in perfect lockstep behind (to
* the left of) the target/parent level traversal.
*/
if (blkno == downlink)
pivotkey_offset = OffsetNumberNext(target_downlinkoffnum);
else
pivotkey_offset = target_downlinkoffnum;
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
topaque = BTPageGetOpaque(state->target);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (!offset_is_negative_infinity(topaque, pivotkey_offset))
{
/*
* If we're looking for the next pivot tuple in target page,
* but there is no more pivot tuples, then we should match to
* high key instead.
*/
if (pivotkey_offset > PageGetMaxOffsetNumber(state->target))
{
if (P_RIGHTMOST(topaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("child high key is greater than rightmost pivot key on target level in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Target block=%u child block=%u target page lsn=%X/%X.",
state->targetblock, blkno,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
pivotkey_offset = P_HIKEY;
}
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, pivotkey_offset);
itup = (IndexTuple) PageGetItem(state->target, itemid);
}
else
{
/*
* We cannot try to match child's high key to a negative
* infinity key in target, since there is nothing to compare.
* However, it's still possible to match child's high key
* outside of target page. The reason why we're are is that
* bt_child_highkey_check() was previously called for the
* cousin page of 'loaded_child', which is incomplete split.
* So, now we traverse to the right of that cousin page and
* current child level page under consideration still belongs
* to the subtree of target's left sibling. Thus, we need to
* match child's high key to it's left uncle page high key.
* Thankfully we saved it, it's called a "low key" of target
* page.
*/
if (!state->lowkey)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("can't find left sibling high key in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Target block=%u child block=%u target page lsn=%X/%X.",
state->targetblock, blkno,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
itup = state->lowkey;
}
Fix amcheck child check pg_upgrade bug. Commit d114cc53 overlooked the fact that pg_upgrade'd B-Tree indexes have leaf page high keys whose offset numbers do not match the one from the copy of the tuple one level up (the copy stored with a downlink for leaf page's right sibling page). This led to false positive reports of corruption from bt_index_parent_check() when it was called to verify a pg_upgrade'd index. To fix, skip comparing the offset number on pg_upgrade'd B-Tree indexes. Author: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Author: Peter Geoghegan <pg@bowt.ie> Reported-By: Andrew Bille <andrewbille@gmail.com> Diagnosed-By: Anastasia Lubennikova <a.lubennikova@postgrespro.ru> Bug: #16619 Discussion: https://postgr.es/m/16619-aaba10f83fdc1c3c@postgresql.org Backpatch: 13-, where child check was enhanced.
2020-09-16 19:42:30 +02:00
if (!bt_pivot_tuple_identical(state->heapkeyspace, highkey, itup))
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
{
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("mismatch between parent key and child high key in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Target block=%u child block=%u target page lsn=%X/%X.",
state->targetblock, blkno,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
}
}
/* Exit if we already found next downlink */
if (blkno == downlink)
{
state->prevrightlink = opaque->btpo_next;
state->previncompletesplit = rightsplit;
return;
}
/* Traverse to the next page using rightlink */
blkno = opaque->btpo_next;
/* Free page contents if it's allocated by us */
if (page != loaded_child)
pfree(page);
first = false;
}
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Checks one of target's downlink against its child page.
*
* Conceptually, the target page continues to be what is checked here. The
* target block is still blamed in the event of finding an invariant violation.
* The downlink insertion into the target is probably where any problem raised
* here arises, and there is no such thing as a parent link, so doing the
* verification this way around is much more practical.
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
*
* This function visits child page and it's sequentially called for each
* downlink of target page. Assuming this we also check downlink connectivity
* here in order to save child page visits.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
static void
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
bt_child_check(BtreeCheckState *state, BTScanInsert targetkey,
OffsetNumber downlinkoffnum)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
ItemId itemid;
IndexTuple itup;
BlockNumber childblock;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
OffsetNumber offset;
OffsetNumber maxoffset;
Page child;
BTPageOpaque copaque;
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
BTPageOpaque topaque;
itemid = PageGetItemIdCareful(state, state->targetblock,
state->target, downlinkoffnum);
itup = (IndexTuple) PageGetItem(state->target, itemid);
childblock = BTreeTupleGetDownLink(itup);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* Caller must have ShareLock on target relation, because of
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* considerations around page deletion by VACUUM.
*
* NB: In general, page deletion deletes the right sibling's downlink, not
* the downlink of the page being deleted; the deleted page's downlink is
* reused for its sibling. The key space is thereby consolidated between
* the deleted page and its right sibling. (We cannot delete a parent
* page's rightmost child unless it is the last child page, and we intend
* to also delete the parent itself.)
*
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
* If this verification happened without a ShareLock, the following race
* condition could cause false positives:
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* In general, concurrent page deletion might occur, including deletion of
* the left sibling of the child page that is examined here. If such a
* page deletion were to occur, closely followed by an insertion into the
* newly expanded key space of the child, a window for the false positive
* opens up: the stale parent/target downlink originally followed to get
* to the child legitimately ceases to be a lower bound on all items in
* the page, since the key space was concurrently expanded "left".
* (Insertion followed the "new" downlink for the child, not our now-stale
* downlink, which was concurrently physically removed in target/parent as
* part of deletion's first phase.)
*
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
* While we use various techniques elsewhere to perform cross-page
* verification for !readonly callers, a similar trick seems difficult
* here. The tricks used by bt_recheck_sibling_links and by
* bt_right_page_check_scankey both involve verification of a same-level,
* cross-sibling invariant. Cross-level invariants are far more squishy,
* though. The nbtree REDO routines do not actually couple buffer locks
* across levels during page splits, so making any cross-level check work
* reliably in !readonly mode may be impossible.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
amcheck: editorialize variable name & comment. No exclusive lock is taken anymore...
2017-03-10 05:01:37 +01:00
Assert(state->readonly);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Verify child page has the downlink key from target page (its parent) as
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* a lower bound; downlink must be strictly less than all keys on the
* page.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Check all items, rather than checking just the first and trusting that
* the operator class obeys the transitive law.
*/
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
topaque = BTPageGetOpaque(state->target);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
child = palloc_btree_page(state, childblock);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
copaque = BTPageGetOpaque(child);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
maxoffset = PageGetMaxOffsetNumber(child);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/*
* Since we've already loaded the child block, combine this check with
* check for downlink connectivity.
*/
bt_child_highkey_check(state, downlinkoffnum,
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
child, topaque->btpo_level);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Since there cannot be a concurrent VACUUM operation in readonly mode,
* and since a page has no links within other pages (siblings and parent)
* once it is marked fully deleted, it should be impossible to land on a
* fully deleted page.
*
* It does not quite make sense to enforce that the page cannot even be
* half-dead, despite the fact the downlink is modified at the same stage
* that the child leaf page is marked half-dead. That's incorrect because
* there may occasionally be multiple downlinks from a chain of pages
* undergoing deletion, where multiple successive calls are made to
* _bt_unlink_halfdead_page() by VACUUM before it can finally safely mark
* the leaf page as fully dead. While _bt_mark_page_halfdead() usually
* removes the downlink to the leaf page that is marked half-dead, that's
* not guaranteed, so it's possible we'll land on a half-dead page with a
* downlink due to an interrupted multi-level page deletion.
*
* We go ahead with our checks if the child page is half-dead. It's safe
* to do so because we do not test the child's high key, so it does not
* matter that the original high key will have been replaced by a dummy
* truncated high key within _bt_mark_page_halfdead(). All other page
* items are left intact on a half-dead page, so there is still something
* to test.
*/
if (P_ISDELETED(copaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("downlink to deleted page found in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Parent block=%u child block=%u parent page lsn=%X/%X.",
state->targetblock, childblock,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
for (offset = P_FIRSTDATAKEY(copaque);
offset <= maxoffset;
offset = OffsetNumberNext(offset))
{
/*
* Skip comparison of target page key against "negative infinity"
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* item, if any. Checking it would indicate that it's not a strict
* lower bound, but that's only because of the hard-coding for
* negative infinity items within _bt_compare().
*
* If nbtree didn't truncate negative infinity tuples during internal
* page splits then we'd expect child's negative infinity key to be
* equal to the scankey/downlink from target/parent (it would be a
* "low key" in this hypothetical scenario, and so it would still need
* to be treated as a special case here).
*
* Negative infinity items can be thought of as a strict lower bound
* that works transitively, with the last non-negative-infinity pivot
* followed during a descent from the root as its "true" strict lower
* bound. Only a small number of negative infinity items are truly
* negative infinity; those that are the first items of leftmost
* internal pages. In more general terms, a negative infinity item is
* only negative infinity with respect to the subtree that the page is
* at the root of.
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
*
* See also: bt_rootdescend(), which can even detect transitive
* inconsistencies on cousin leaf pages.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
if (offset_is_negative_infinity(copaque, offset))
continue;
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
if (!invariant_l_nontarget_offset(state, targetkey, childblock, child,
offset))
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("down-link lower bound invariant violated for index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Parent block=%u child index tid=(%u,%u) parent page lsn=%X/%X.",
state->targetblock, childblock, offset,
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
LSN_FORMAT_ARGS(state->targetlsn))));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
pfree(child);
}
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Checks if page is missing a downlink that it should have.
*
* A page that lacks a downlink/parent may indicate corruption. However, we
* must account for the fact that a missing downlink can occasionally be
* encountered in a non-corrupt index. This can be due to an interrupted page
* split, or an interrupted multi-level page deletion (i.e. there was a hard
* crash or an error during a page split, or while VACUUM was deleting a
* multi-level chain of pages).
*
* Note that this can only be called in readonly mode, so there is no need to
* be concerned about concurrent page splits or page deletions.
*/
static void
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
BlockNumber blkno, Page page)
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
{
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
BTPageOpaque opaque = BTPageGetOpaque(page);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ItemId itemid;
IndexTuple itup;
Page child;
BTPageOpaque copaque;
uint32 level;
BlockNumber childblk;
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
XLogRecPtr pagelsn;
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
Assert(state->readonly);
Assert(!P_IGNORE(opaque));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/* No next level up with downlinks to fingerprint from the true root */
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (P_ISROOT(opaque))
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
return;
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
pagelsn = PageGetLSN(page);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Incomplete (interrupted) page splits can account for the lack of a
* downlink. Some inserting transaction should eventually complete the
* page split in passing, when it notices that the left sibling page is
* P_INCOMPLETE_SPLIT().
*
* In general, VACUUM is not prepared for there to be no downlink to a
* page that it deletes. This is the main reason why the lack of a
* downlink can be reported as corruption here. It's not obvious that an
* invalid missing downlink can result in wrong answers to queries,
* though, since index scans that land on the child may end up
* consistently moving right. The handling of concurrent page splits (and
* page deletions) within _bt_moveright() cannot distinguish
* inconsistencies that last for a moment from inconsistencies that are
* permanent and irrecoverable.
*
* VACUUM isn't even prepared to delete pages that have no downlink due to
* an incomplete page split, but it can detect and reason about that case
* by design, so it shouldn't be taken to indicate corruption. See
* _bt_pagedel() for full details.
*/
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (rightsplit)
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
{
ereport(DEBUG1,
(errcode(ERRCODE_NO_DATA),
Improve quoting in some error messages
2021-04-14 09:10:57 +02:00
errmsg_internal("harmless interrupted page split detected in index \"%s\"",
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u level=%u left sibling=%u page lsn=%X/%X.",
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
blkno, opaque->btpo_level,
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
opaque->btpo_prev,
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
LSN_FORMAT_ARGS(pagelsn))));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
return;
}
/*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* Page under check is probably the "top parent" of a multi-level page
* deletion. We'll need to descend the subtree to make sure that
* descendant pages are consistent with that, though.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* If the page (which must be non-ignorable) is a leaf page, then clearly
* it can't be the top parent. The lack of a downlink is probably a
* symptom of a broad problem that could just as easily cause
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
* inconsistencies anywhere else.
*/
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (P_ISLEAF(opaque))
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("leaf index block lacks downlink in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u page lsn=%X/%X.",
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
blkno,
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
LSN_FORMAT_ARGS(pagelsn))));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
/* Descend from the given page, which is an internal page */
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
elog(DEBUG1, "checking for interrupted multi-level deletion due to missing downlink in index \"%s\"",
RelationGetRelationName(state->rel));
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
level = opaque->btpo_level;
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
itemid = PageGetItemIdCareful(state, blkno, page, P_FIRSTDATAKEY(opaque));
itup = (IndexTuple) PageGetItem(page, itemid);
Rename nbtree tuple macros. Rename two function-style macros, removing the word "inner". This makes things more consistent.
2019-12-17 02:49:45 +01:00
childblk = BTreeTupleGetDownLink(itup);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
for (;;)
{
CHECK_FOR_INTERRUPTS();
child = palloc_btree_page(state, childblk);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
copaque = BTPageGetOpaque(child);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
if (P_ISLEAF(copaque))
break;
/* Do an extra sanity check in passing on internal pages */
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
if (copaque->btpo_level != level - 1)
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("downlink points to block in index \"%s\" whose level is not one level down",
RelationGetRelationName(state->rel)),
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
errdetail_internal("Top parent/under check block=%u block pointed to=%u expected level=%u level in pointed to block=%u.",
blkno, childblk,
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
level - 1, copaque->btpo_level)));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
level = copaque->btpo_level;
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
itemid = PageGetItemIdCareful(state, childblk, child,
P_FIRSTDATAKEY(copaque));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
itup = (IndexTuple) PageGetItem(child, itemid);
Rename nbtree tuple macros. Rename two function-style macros, removing the word "inner". This makes things more consistent.
2019-12-17 02:49:45 +01:00
childblk = BTreeTupleGetDownLink(itup);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/* Be slightly more pro-active in freeing this memory, just in case */
pfree(child);
}
/*
* Since there cannot be a concurrent VACUUM operation in readonly mode,
* and since a page has no links within other pages (siblings and parent)
* once it is marked fully deleted, it should be impossible to land on a
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* fully deleted page. See bt_child_check() for further details.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* The bt_child_check() P_ISDELETED() check is repeated here because
* bt_child_check() does not visit pages reachable through negative
* infinity items. Besides, bt_child_check() is unwilling to descend
* multiple levels. (The similar bt_child_check() P_ISDELETED() check
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
* within bt_check_level_from_leftmost() won't reach the page either,
Add amcheck missing downlink tests. Also use palloc0() for main amcheck state, and adjust a few comments. Somehow I pushed old version of patch in commit 4eaf7eaccb291f5d32d9f05284bfca9c5744de9a, so commit the difference. Peter Geoghegan
2018-04-25 21:05:53 +02:00
* since the leaf's live siblings should have their sibling links updated
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
* to bypass the deletion target page when it is marked fully dead.)
*
* If this error is raised, it might be due to a previous multi-level page
* deletion that failed to realize that it wasn't yet safe to mark the
* leaf page as fully dead. A "dangling downlink" will still remain when
* this happens. The fact that the dangling downlink's page (the leaf's
* parent/ancestor page) lacked a downlink is incidental.
*/
if (P_ISDELETED(copaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("downlink to deleted leaf page found in index \"%s\"",
RelationGetRelationName(state->rel)),
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
errdetail_internal("Top parent/target block=%u leaf block=%u top parent/under check lsn=%X/%X.",
blkno, childblk,
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
LSN_FORMAT_ARGS(pagelsn))));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Iff leaf page is half-dead, its high key top parent link should point
* to what VACUUM considered to be the top parent page at the instant it
* was interrupted. Provided the high key link actually points to the
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* page under check, the missing downlink we detected is consistent with
* there having been an interrupted multi-level page deletion. This means
* that the subtree with the page under check at its root (a page deletion
* chain) is in a consistent state, enabling VACUUM to resume deleting the
* entire chain the next time it encounters the half-dead leaf page.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*/
if (P_ISHALFDEAD(copaque) && !P_RIGHTMOST(copaque))
{
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
itemid = PageGetItemIdCareful(state, childblk, child, P_HIKEY);
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
itup = (IndexTuple) PageGetItem(child, itemid);
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
if (BTreeTupleGetTopParent(itup) == blkno)
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
return;
}
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("internal index block lacks downlink in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Block=%u level=%u page lsn=%X/%X.",
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
blkno, opaque->btpo_level,
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
LSN_FORMAT_ARGS(pagelsn))));
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
}
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/*
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* Per-tuple callback from table_index_build_scan, used to determine if index has
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* all the entries that definitely should have been observed in leaf pages of
* the target index (that is, all IndexTuples that were fingerprinted by our
* Bloom filter). All heapallindexed checks occur here.
*
* The redundancy between an index and the table it indexes provides a good
* opportunity to detect corruption, especially corruption within the table.
* The high level principle behind the verification performed here is that any
* IndexTuple that should be in an index following a fresh CREATE INDEX (based
* on the same index definition) should also have been in the original,
* existing index, which should have used exactly the same representation
*
* Since the overall structure of the index has already been verified, the most
* likely explanation for error here is a corrupt heap page (could be logical
* or physical corruption). Index corruption may still be detected here,
* though. Only readonly callers will have verified that left links and right
* links are in agreement, and so it's possible that a leaf page transposition
* within index is actually the source of corruption detected here (for
* !readonly callers). The checks performed only for readonly callers might
* more accurately frame the problem as a cross-page invariant issue (this
* could even be due to recovery not replaying all WAL records). The !readonly
* ERROR message raised here includes a HINT about retrying with readonly
* verification, just in case it's a cross-page invariant issue, though that
* isn't particularly likely.
*
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* table_index_build_scan() expects to be able to find the root tuple when a
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* heap-only tuple (the live tuple at the end of some HOT chain) needs to be
* indexed, in order to replace the actual tuple's TID with the root tuple's
* TID (which is what we're actually passed back here). The index build heap
* scan code will raise an error when a tuple that claims to be the root of the
* heap-only tuple's HOT chain cannot be located. This catches cases where the
* original root item offset/root tuple for a HOT chain indicates (for whatever
* reason) that the entire HOT chain is dead, despite the fact that the latest
* heap-only tuple should be indexed. When this happens, sequential scans may
* always give correct answers, and all indexes may be considered structurally
* consistent (i.e. the nbtree structural checks would not detect corruption).
* It may be the case that only index scans give wrong answers, and yet heap or
* SLRU corruption is the real culprit. (While it's true that LP_DEAD bit
* setting will probably also leave the index in a corrupt state before too
* long, the problem is nonetheless that there is heap corruption.)
*
tableam: Support for an index build's initial table scan(s). To support building indexes over tables of different AMs, the scans to do so need to be routed through the table AM. While moving a fair amount of code, nearly all the changes are just moving code to below a callback. Currently the range based interface wouldn't make much sense for non block based table AMs. But that seems aceptable for now. Author: Andres Freund Discussion: https://postgr.es/m/20180703070645.wchpu5muyto5n647@alap3.anarazel.de
2019-03-28 03:59:06 +01:00
* Heap-only tuple handling within table_index_build_scan() works in a way that
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
* helps us to detect index tuples that contain the wrong values (values that
* don't match the latest tuple in the HOT chain). This can happen when there
* is no superseding index tuple due to a faulty assessment of HOT safety,
* perhaps during the original CREATE INDEX. Because the latest tuple's
* contents are used with the root TID, an error will be raised when a tuple
* with the same TID but non-matching attribute values is passed back to us.
* Faulty assessment of HOT-safety was behind at least two distinct CREATE
* INDEX CONCURRENTLY bugs that made it into stable releases, one of which was
* undetected for many years. In short, the same principle that allows a
* REINDEX to repair corruption when there was an (undetected) broken HOT chain
* also allows us to detect the corruption in many cases.
*/
static void
Pass ItemPointer not HeapTuple to IndexBuildCallback. Not all AMs use HeapTuples internally, making it inconvenient to pass a HeapTuple. As the index callbacks really only need the TID, not the full tuple, modify callback to only take ItemPointer. Author: Ashwin Agrawal Reviewed-By: Andres Freund Discussion: https://postgr.es/m/CALfoeis6=8ehuR=VNtHvj3z16cYfCwPdTcpaxU+sfSUJ5QgR3g@mail.gmail.com
2019-11-08 09:44:52 +01:00
bt_tuple_present_callback(Relation index, ItemPointer tid, Datum *values,
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
bool *isnull, bool tupleIsAlive, void *checkstate)
{
BtreeCheckState *state = (BtreeCheckState *) checkstate;
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
IndexTuple itup,
norm;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
Assert(state->heapallindexed);
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
/* Generate a normalized index tuple for fingerprinting */
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
itup = index_form_tuple(RelationGetDescr(index), values, isnull);
Pass ItemPointer not HeapTuple to IndexBuildCallback. Not all AMs use HeapTuples internally, making it inconvenient to pass a HeapTuple. As the index callbacks really only need the TID, not the full tuple, modify callback to only take ItemPointer. Author: Ashwin Agrawal Reviewed-By: Andres Freund Discussion: https://postgr.es/m/CALfoeis6=8ehuR=VNtHvj3z16cYfCwPdTcpaxU+sfSUJ5QgR3g@mail.gmail.com
2019-11-08 09:44:52 +01:00
itup->t_tid = *tid;
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
norm = bt_normalize_tuple(state, itup);
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
/* Probe Bloom filter -- tuple should be present */
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
if (bloom_lacks_element(state->filter, (unsigned char *) norm,
IndexTupleSize(norm)))
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("heap tuple (%u,%u) from table \"%s\" lacks matching index tuple within index \"%s\"",
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ItemPointerGetBlockNumber(&(itup->t_tid)),
ItemPointerGetOffsetNumber(&(itup->t_tid)),
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
RelationGetRelationName(state->heaprel),
RelationGetRelationName(state->rel)),
!state->readonly
? errhint("Retrying verification using the function bt_index_parent_check() might provide a more specific error.")
: 0));
state->heaptuplespresent++;
pfree(itup);
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
/* Cannot leak memory here */
if (norm != itup)
pfree(norm);
}
/*
* Normalize an index tuple for fingerprinting.
*
* In general, index tuple formation is assumed to be deterministic by
* heapallindexed verification, and IndexTuples are assumed immutable. While
* the LP_DEAD bit is mutable in leaf pages, that's ItemId metadata, which is
* not fingerprinted. Normalization is required to compensate for corner
* cases where the determinism assumption doesn't quite work.
*
* There is currently one such case: index_form_tuple() does not try to hide
* the source TOAST state of input datums. The executor applies TOAST
* compression for heap tuples based on different criteria to the compression
* applied within btinsert()'s call to index_form_tuple(): it sometimes
* compresses more aggressively, resulting in compressed heap tuple datums but
* uncompressed corresponding index tuple datums. A subsequent heapallindexed
* verification will get a logically equivalent though bitwise unequal tuple
* from index_form_tuple(). False positive heapallindexed corruption reports
* could occur without normalizing away the inconsistency.
*
* Returned tuple is often caller's own original tuple. Otherwise, it is a
* new representation of caller's original index tuple, palloc()'d in caller's
* memory context.
*
* Note: This routine is not concerned with distinctions about the
* representation of tuples beyond those that might break heapallindexed
* verification. In particular, it won't try to normalize opclass-equal
* datums with potentially distinct representations (e.g., btree/numeric_ops
* index datums will not get their display scale normalized-away here).
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
* Caller does normalization for non-pivot tuples that have a posting list,
* since dummy CREATE INDEX callback code generates new tuples with the same
* normalized representation.
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
*/
static IndexTuple
bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
{
TupleDesc tupleDescriptor = RelationGetDescr(state->rel);
Datum normalized[INDEX_MAX_KEYS];
bool isnull[INDEX_MAX_KEYS];
bool toast_free[INDEX_MAX_KEYS];
bool formnewtup = false;
IndexTuple reformed;
int i;
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/* Caller should only pass "logical" non-pivot tuples here */
Assert(!BTreeTupleIsPosting(itup) && !BTreeTupleIsPivot(itup));
Avoid amcheck inline compression false positives. The previous tacit assumption that index_form_tuple() hides differences in the TOAST state of its input datums was wrong. Normalize input varlena datums by decompressing compressed values, and forming a new index tuple for fingerprinting using uncompressed inputs. The final normalized representation may actually be compressed once again within index_form_tuple(), though that shouldn't matter. When the original tuple is found to have no datums that are compressed inline, fingerprint the original tuple directly. Normalization avoids false positive reports of corruption in certain cases. For example, the executor can apply toasting with some inline compression to an entire heap tuple because its input has a single external TOAST pointer. Varlena datums for other attributes that are not particularly good candidates for inline compression can be compressed in the heap tuple in passing, without the representation of the same values in index tuples ever receiving concomitant inline compression. Add a test case to recreate the issue in a simpler though less realistic way: by exploiting differences in pg_attribute.attstorage between heap and index relations. This bug was discovered by me during testing of an upcoming set of nbtree enhancements. It was also independently reported by Andreas Kunert, as bug #15597. His test case was rather more realistic than the one I ended up using. Bug: #15597 Discussion: https://postgr.es/m/CAH2-WznrVd9ie+TTJ45nDT+v2nUt6YJwQrT9SebCdQKtAvfPZw@mail.gmail.com Discussion: https://postgr.es/m/15597-294e5d3e7f01c407@postgresql.org Backpatch: 11-, where heapallindexed verification was introduced.
2019-02-07 00:54:19 +01:00
/* Easy case: It's immediately clear that tuple has no varlena datums */
if (!IndexTupleHasVarwidths(itup))
return itup;
for (i = 0; i < tupleDescriptor->natts; i++)
{
Form_pg_attribute att;
att = TupleDescAttr(tupleDescriptor, i);
/* Assume untoasted/already normalized datum initially */
toast_free[i] = false;
normalized[i] = index_getattr(itup, att->attnum,
tupleDescriptor,
&isnull[i]);
if (att->attbyval || att->attlen != -1 || isnull[i])
continue;
/*
* Callers always pass a tuple that could safely be inserted into the
* index without further processing, so an external varlena header
* should never be encountered here
*/
if (VARATT_IS_EXTERNAL(DatumGetPointer(normalized[i])))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("external varlena datum in tuple that references heap row (%u,%u) in index \"%s\"",
ItemPointerGetBlockNumber(&(itup->t_tid)),
ItemPointerGetOffsetNumber(&(itup->t_tid)),
RelationGetRelationName(state->rel))));
else if (VARATT_IS_COMPRESSED(DatumGetPointer(normalized[i])))
{
formnewtup = true;
normalized[i] = PointerGetDatum(PG_DETOAST_DATUM(normalized[i]));
toast_free[i] = true;
}
}
/* Easier case: Tuple has varlena datums, none of which are compressed */
if (!formnewtup)
return itup;
/*
* Hard case: Tuple had compressed varlena datums that necessitate
* creating normalized version of the tuple from uncompressed input datums
* (normalized input datums). This is rather naive, but shouldn't be
* necessary too often.
*
* Note that we rely on deterministic index_form_tuple() TOAST compression
* of normalized input.
*/
reformed = index_form_tuple(tupleDescriptor, normalized, isnull);
reformed->t_tid = itup->t_tid;
/* Cannot leak memory here */
for (i = 0; i < tupleDescriptor->natts; i++)
if (toast_free[i])
pfree(DatumGetPointer(normalized[i]));
return reformed;
Add amcheck verification of heap relations belonging to btree indexes. Add a new, optional, capability to bt_index_check() and bt_index_parent_check(): check that each heap tuple that should have an index entry does in fact have one. The extra checking is performed at the end of the existing nbtree checks. This is implemented by using a Bloom filter data structure. The implementation performs set membership tests within a callback (the same type of callback that each index AM registers for CREATE INDEX). The Bloom filter is populated during the initial index verification scan. Reusing the CREATE INDEX infrastructure allows the new verification option to automatically benefit from the heap consistency checks that CREATE INDEX already performs. CREATE INDEX does thorough sanity checking of HOT chains, so the new check actually manages to detect problems in heap-only tuples. Author: Peter Geoghegan Reviewed-By: Pavan Deolasee, Andres Freund Discussion: https://postgr.es/m/CAH2-Wzm5VmG7cu1N-H=nnS57wZThoSDQU+F5dewx3o84M+jY=g@mail.gmail.com
2018-04-01 04:52:01 +02:00
}
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/*
* Produce palloc()'d "plain" tuple for nth posting list entry/TID.
*
* In general, deduplication is not supposed to change the logical contents of
* an index. Multiple index tuples are merged together into one equivalent
* posting list index tuple when convenient.
*
* heapallindexed verification must normalize-away this variation in
* representation by converting posting list tuples into two or more "plain"
* tuples. Each tuple must be fingerprinted separately -- there must be one
* tuple for each corresponding Bloom filter probe during the heap scan.
*
* Note: Caller still needs to call bt_normalize_tuple() with returned tuple.
*/
static inline IndexTuple
bt_posting_plain_tuple(IndexTuple itup, int n)
{
Assert(BTreeTupleIsPosting(itup));
/* Returns non-posting-list tuple */
return _bt_form_posting(itup, BTreeTupleGetPostingN(itup, n), 1);
}
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
/*
* Search for itup in index, starting from fast root page. itup must be a
* non-pivot tuple. This is only supported with heapkeyspace indexes, since
Tweak some nbtree-related code comments.
2019-03-29 20:29:05 +01:00
* we rely on having fully unique keys to find a match with only a single
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
* visit to a leaf page, barring an interrupted page split, where we may have
* to move right. (A concurrent page split is impossible because caller must
* be readonly caller.)
*
* This routine can detect very subtle transitive consistency issues across
* more than one level of the tree. Leaf pages all have a high key (even the
* rightmost page has a conceptual positive infinity high key), but not a low
* key. Their downlink in parent is a lower bound, which along with the high
* key is almost enough to detect every possible inconsistency. A downlink
* separator key value won't always be available from parent, though, because
* the first items of internal pages are negative infinity items, truncated
* down to zero attributes during internal page splits. While it's true that
Improve checking of child pages in contrib/amcheck. This commit eliminates lossiness in check for missing parent downlinks in B-tree. Instead of collecting lossy bitmap, we check for missing downlinks while visiting child pages referenced by downlinks of target level. We traverse from previous child page to the subsequent child page by right links. Intermediate pages are candidates to have lost parent downlinks. Also this commit introduces matching of child high key to the pivot key of it's parent. Discussion: https://postgr.es/m/CAPpHfduoF-c4RhOyOm%3D4-Y367%2B8txq9Q6iM_ty0OYc8si1Abww%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-03-11 10:00:31 +01:00
* bt_child_check() and the high key check can detect most imaginable key
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
* space problems, there are remaining problems it won't detect with non-pivot
* tuples in cousin leaf pages. Starting a search from the root for every
* existing leaf tuple detects small inconsistencies in upper levels of the
* tree that cannot be detected any other way. (Besides all this, this is
* probably also useful as a direct test of the code used by index scans
* themselves.)
*/
static bool
bt_rootdescend(BtreeCheckState *state, IndexTuple itup)
{
BTScanInsert key;
BTStack stack;
Buffer lbuf;
bool exists;
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
key = _bt_mkscankey(state->rel, state->heaprel, itup);
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
Assert(key->heapkeyspace && key->scantid != NULL);
/*
* Search from root.
*
* Ideally, we would arrange to only move right within _bt_search() when
* an interrupted page split is detected (i.e. when the incomplete split
* bit is found to be set), but for now we accept the possibility that
* that could conceal an inconsistency.
*/
Assert(state->readonly && state->rootdescend);
exists = false;
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
stack = _bt_search(state->rel, state->heaprel, key, &lbuf, BT_READ, NULL);
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
if (BufferIsValid(lbuf))
{
BTInsertStateData insertstate;
OffsetNumber offnum;
Page page;
insertstate.itup = itup;
insertstate.itemsz = MAXALIGN(IndexTupleSize(itup));
insertstate.itup_key = key;
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
insertstate.postingoff = 0;
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
insertstate.bounds_valid = false;
insertstate.buf = lbuf;
/* Get matching tuple on leaf page */
offnum = _bt_binsrch_insert(state->rel, &insertstate);
/* Compare first >= matching item on leaf page, if any */
page = BufferGetPage(lbuf);
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/* Should match on first heap TID when tuple has a posting list */
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
if (offnum <= PageGetMaxOffsetNumber(page) &&
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
insertstate.postingoff <= 0 &&
Allow amcheck to re-find tuples using new search. Teach contrib/amcheck's bt_index_parent_check() function to take advantage of the uniqueness property of heapkeyspace indexes in support of a new verification option: non-pivot tuples (non-highkey tuples on the leaf level) can optionally be re-found using a new search for each, that starts from the root page. If a tuple cannot be re-found, report that the index is corrupt. The new "rootdescend" verification option is exhaustive, and can therefore make a call to bt_index_parent_check() take a lot longer. Re-finding tuples during verification is mostly intended as an option for backend developers, since the corruption scenarios that it alone is uniquely capable of detecting seem fairly far-fetched. For example, "rootdescend" verification is much more likely to detect corruption of the least significant byte of a key from a pivot tuple in the root page of a B-Tree that already has at least three levels. Typically, only a few tuples on a cousin leaf page are at risk of "getting overlooked" by index scans in this scenario. The corrupt key in the root page is only slightly corrupt: corrupt enough to give wrong answers to some queries, and yet not corrupt enough to allow the problem to be detected without verifying agreement between the leaf page and the root page, skipping at least one internal page level. The existing bt_index_parent_check() checks never cross more than a single level. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
2019-03-20 18:41:36 +01:00
_bt_compare(state->rel, key, page, offnum) == 0)
exists = true;
_bt_relbuf(state->rel, lbuf);
}
_bt_freestack(stack);
pfree(key);
return exists;
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Is particular offset within page (whose special state is passed by caller)
* the page negative-infinity item?
*
* As noted in comments above _bt_compare(), there is special handling of the
* first data item as a "negative infinity" item. The hard-coding within
* _bt_compare() makes comparing this item for the purposes of verification
* pointless at best, since the IndexTuple only contains a valid TID (a
* reference TID to child page).
*/
static inline bool
offset_is_negative_infinity(BTPageOpaque opaque, OffsetNumber offset)
{
/*
* For internal pages only, the first item after high key, if any, is
* negative infinity item. Internal pages always have a negative infinity
* item, whereas leaf pages never have one. This implies that negative
* infinity item is either first or second line item, or there is none
* within page.
*
Adjust INCLUDE index truncation comments and code. Add several assertions that ensure that we're dealing with a pivot tuple without non-key attributes where that's expected. Also, remove the assertion within _bt_isequal(), restoring the v10 function signature. A similar check will be performed for the page highkey within _bt_moveright() in most cases. Also avoid dropping all objects within regression tests, to increase pg_dump test coverage for INCLUDE indexes. Rather than using infrastructure that's generally intended to be used with reference counted heap tuple descriptors during truncation, use the same function that was introduced to store flat TupleDescs in shared memory (we use a temp palloc'd buffer). This isn't strictly necessary, but seems more future-proof than the old approach. It also lets us avoid including rel.h within indextuple.c, which was arguably a modularity violation. Also, we now call index_deform_tuple() with the truncated TupleDesc, not the source TupleDesc, since that's more robust, and saves a few cycles. In passing, fix a memory leak by pfree'ing truncated pivot tuple memory during CREATE INDEX. Also pfree during a page split, just to be consistent. Refactor _bt_check_natts() to be more readable. Author: Peter Geoghegan with some editorization by me Reviewed by: Alexander Korotkov, Teodor Sigaev Discussion: https://www.postgresql.org/message-id/CAH2-Wz%3DkCWuXeMrBCopC-tFs3FbiVxQNjjgNKdG2sHxZ5k2y3w%40mail.gmail.com
2018-04-19 07:45:58 +02:00
* Negative infinity items are a special case among pivot tuples. They
* always have zero attributes, while all other pivot tuples always have
* nkeyatts attributes.
Indexes with INCLUDE columns and their support in B-tree This patch introduces INCLUDE clause to index definition. This clause specifies a list of columns which will be included as a non-key part in the index. The INCLUDE columns exist solely to allow more queries to benefit from index-only scans. Also, such columns don't need to have appropriate operator classes. Expressions are not supported as INCLUDE columns since they cannot be used in index-only scans. Index access methods supporting INCLUDE are indicated by amcaninclude flag in IndexAmRoutine. For now, only B-tree indexes support INCLUDE clause. In B-tree indexes INCLUDE columns are truncated from pivot index tuples (tuples located in non-leaf pages and high keys). Therefore, B-tree indexes now might have variable number of attributes. This patch also provides generic facility to support that: pivot tuples contain number of their attributes in t_tid.ip_posid. Free 13th bit of t_info is used for indicating that. This facility will simplify further support of index suffix truncation. The changes of above are backward-compatible, pg_upgrade doesn't need special handling of B-tree indexes for that. Bump catalog version Author: Anastasia Lubennikova with contribition by Alexander Korotkov and me Reviewed by: Peter Geoghegan, Tomas Vondra, Antonin Houska, Jeff Janes, David Rowley, Alexander Korotkov Discussion: https://www.postgresql.org/message-id/flat/56168952.4010101@postgrespro.ru
2018-04-07 22:00:39 +02:00
*
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* Right-most pages don't have a high key, but could be said to
* conceptually have a "positive infinity" high key. Thus, there is a
* symmetry between down link items in parent pages, and high keys in
* children. Together, they represent the part of the key space that
* belongs to each page in the index. For example, all children of the
* root page will have negative infinity as a lower bound from root
* negative infinity downlink, and positive infinity as an upper bound
* (implicitly, from "imaginary" positive infinity high key in root).
*/
return !P_ISLEAF(opaque) && offset == P_FIRSTDATAKEY(opaque);
}
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/*
* Does the invariant hold that the key is strictly less than a given upper
* bound offset item?
*
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
* Verifies line pointer on behalf of caller.
*
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* If this function returns false, convention is that caller throws error due
* to corruption.
*/
static inline bool
invariant_l_offset(BtreeCheckState *state, BTScanInsert key,
OffsetNumber upperbound)
{
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
ItemId itemid;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
int32 cmp;
Assert(key->pivotsearch);
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
/* Verify line pointer before checking tuple */
itemid = PageGetItemIdCareful(state, state->targetblock, state->target,
upperbound);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/* pg_upgrade'd indexes may legally have equal sibling tuples */
if (!key->heapkeyspace)
return invariant_leq_offset(state, key, upperbound);
cmp = _bt_compare(state->rel, key, state->target, upperbound);
/*
* _bt_compare() is capable of determining that a scankey with a
* filled-out attribute is greater than pivot tuples where the comparison
* is resolved at a truncated attribute (value of attribute in pivot is
* minus infinity). However, it is not capable of determining that a
* scankey is _less than_ a tuple on the basis of a comparison resolved at
* _scankey_ minus infinity attribute. Complete an extra step to simulate
* having minus infinity values for omitted scankey attribute(s).
*/
if (cmp == 0)
{
BTPageOpaque topaque;
IndexTuple ritup;
int uppnkeyatts;
ItemPointer rheaptid;
bool nonpivot;
ritup = (IndexTuple) PageGetItem(state->target, itemid);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
topaque = BTPageGetOpaque(state->target);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
nonpivot = P_ISLEAF(topaque) && upperbound >= P_FIRSTDATAKEY(topaque);
/* Get number of keys + heap TID for item to the right */
uppnkeyatts = BTreeTupleGetNKeyAtts(ritup, state->rel);
rheaptid = BTreeTupleGetHeapTIDCareful(state, ritup, nonpivot);
/* Heap TID is tiebreaker key attribute */
if (key->keysz == uppnkeyatts)
return key->scantid == NULL && rheaptid != NULL;
return key->keysz < uppnkeyatts;
}
return cmp < 0;
}
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
/*
* Does the invariant hold that the key is less than or equal to a given upper
* bound offset item?
*
Standardize ItemIdData terminology. The term "item pointer" should not be used to refer to ItemIdData variables, since that is needlessly ambiguous. Only ItemPointerData/ItemPointer variables should be called item pointers. To fix, establish the convention that ItemIdData variables should always be referred to either as "item identifiers" or "line pointers". The term "item identifier" already predominates in docs and translatable messages, and so should be the preferred alternative there. Discussion: https://postgr.es/m/CAH2-Wz=c=MZQjUzde3o9+2PLAPuHTpVZPPdYxN=E4ndQ2--8ew@mail.gmail.com
2019-05-14 00:53:39 +02:00
* Caller should have verified that upperbound's line pointer is consistent
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
* using PageGetItemIdCareful() call.
*
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* If this function returns false, convention is that caller throws error due
* to corruption.
*/
static inline bool
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
invariant_leq_offset(BtreeCheckState *state, BTScanInsert key,
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
OffsetNumber upperbound)
{
int32 cmp;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Assert(key->pivotsearch);
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
cmp = _bt_compare(state->rel, key, state->target, upperbound);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
return cmp <= 0;
}
/*
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* Does the invariant hold that the key is strictly greater than a given lower
* bound offset item?
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
Standardize ItemIdData terminology. The term "item pointer" should not be used to refer to ItemIdData variables, since that is needlessly ambiguous. Only ItemPointerData/ItemPointer variables should be called item pointers. To fix, establish the convention that ItemIdData variables should always be referred to either as "item identifiers" or "line pointers". The term "item identifier" already predominates in docs and translatable messages, and so should be the preferred alternative there. Discussion: https://postgr.es/m/CAH2-Wz=c=MZQjUzde3o9+2PLAPuHTpVZPPdYxN=E4ndQ2--8ew@mail.gmail.com
2019-05-14 00:53:39 +02:00
* Caller should have verified that lowerbound's line pointer is consistent
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
* using PageGetItemIdCareful() call.
*
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* If this function returns false, convention is that caller throws error due
* to corruption.
*/
static inline bool
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
invariant_g_offset(BtreeCheckState *state, BTScanInsert key,
OffsetNumber lowerbound)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
int32 cmp;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Assert(key->pivotsearch);
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
cmp = _bt_compare(state->rel, key, state->target, lowerbound);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/* pg_upgrade'd indexes may legally have equal sibling tuples */
if (!key->heapkeyspace)
return cmp >= 0;
/*
* No need to consider the possibility that scankey has attributes that we
* need to force to be interpreted as negative infinity. _bt_compare() is
* able to determine that scankey is greater than negative infinity. The
* distinction between "==" and "<" isn't interesting here, since
* corruption is indicated either way.
*/
return cmp > 0;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* Does the invariant hold that the key is strictly less than a given upper
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
* bound offset item, with the offset relating to a caller-supplied page that
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
* is not the current target page?
*
* Caller's non-target page is a child page of the target, checked as part of
* checking a property of the target page (i.e. the key comes from the
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
* target). Verifies line pointer on behalf of caller.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* If this function returns false, convention is that caller throws error due
* to corruption.
*/
static inline bool
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
invariant_l_nontarget_offset(BtreeCheckState *state, BTScanInsert key,
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
BlockNumber nontargetblock, Page nontarget,
OffsetNumber upperbound)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
{
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
ItemId itemid;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
int32 cmp;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Assert(key->pivotsearch);
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
/* Verify line pointer before checking tuple */
itemid = PageGetItemIdCareful(state, nontargetblock, nontarget,
upperbound);
Refactor nbtree insertion scankeys. Use dedicated struct to represent nbtree insertion scan keys. Having a dedicated struct makes the difference between search type scankeys and insertion scankeys a lot clearer, and simplifies the signature of several related functions. This is based on a suggestion by Andrey Lepikhov. Streamline how unique index insertions cache binary search progress. Cache the state of in-progress binary searches within _bt_check_unique() for later instead of having callers avoid repeating the binary search in an ad-hoc manner. This makes it easy to add a new optimization: _bt_check_unique() now falls out of its loop immediately in the common case where it's already clear that there couldn't possibly be a duplicate. The new _bt_check_unique() scheme makes it a lot easier to manage cached binary search effort afterwards, from within _bt_findinsertloc(). This is needed for the upcoming patch to make nbtree tuples unique by treating heap TID as a final tiebreaker column. Unique key binary searches need to restore lower and upper bounds. They cannot simply continue to use the >= lower bound as the offset to insert at, because the heap TID tiebreaker column must be used in comparisons for the restored binary search (unlike the original _bt_check_unique() binary search, where scankey's heap TID column must be omitted). Author: Peter Geoghegan, Heikki Linnakangas Reviewed-By: Heikki Linnakangas, Andrey Lepikhov Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
2019-03-20 17:30:57 +01:00
cmp = _bt_compare(state->rel, key, nontarget, upperbound);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/* pg_upgrade'd indexes may legally have equal sibling tuples */
if (!key->heapkeyspace)
return cmp <= 0;
/* See invariant_l_offset() for an explanation of this extra step */
if (cmp == 0)
{
IndexTuple child;
int uppnkeyatts;
ItemPointer childheaptid;
BTPageOpaque copaque;
bool nonpivot;
child = (IndexTuple) PageGetItem(nontarget, itemid);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
copaque = BTPageGetOpaque(nontarget);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
nonpivot = P_ISLEAF(copaque) && upperbound >= P_FIRSTDATAKEY(copaque);
/* Get number of keys + heap TID for child/non-target item */
uppnkeyatts = BTreeTupleGetNKeyAtts(child, state->rel);
childheaptid = BTreeTupleGetHeapTIDCareful(state, child, nonpivot);
/* Heap TID is tiebreaker key attribute */
if (key->keysz == uppnkeyatts)
return key->scantid == NULL && childheaptid != NULL;
return key->keysz < uppnkeyatts;
}
return cmp < 0;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
* Given a block number of a B-Tree page, return page in palloc()'d memory.
* While at it, perform some basic checks of the page.
*
* There is never an attempt to get a consistent view of multiple pages using
* multiple concurrent buffer locks; in general, we only acquire a single pin
* and buffer lock at a time, which is often all that the nbtree code requires.
Teach amcheck to verify sibling links in all cases. Teach contrib/amcheck's bt_index_check() function to check agreement between siblings links. The left sibling's right link should point to a right sibling page whose left link points back to the same original left sibling. This extends a check that bt_index_parent_check() always performed to bt_index_check(). This is the first time amcheck has been taught to perform buffer lock coupling, which we have explicitly avoided up until now. The sibling link check tends to catch a lot of real world index corruption with little overhead, so it seems worth accepting the complexity. Note that the new lock coupling logic would not work correctly on replica servers without the changes made by commits 0a7d771f and 9a9db08a (there could be false positives without those changes). Author: Andrey Borodin, Peter Geoghegan Discussion: https://postgr.es/m/0EB0CFA8-CBD8-4296-8049-A2C0F28FAE8C@yandex-team.ru
2020-08-08 20:12:01 +02:00
* (Actually, bt_recheck_sibling_links couples buffer locks, which is the only
* exception to this general rule.)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*
* Operating on a copy of the page is useful because it prevents control
* getting stuck in an uninterruptible state when an underlying operator class
* misbehaves.
*/
static Page
palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum)
{
Buffer buffer;
Page page;
BTPageOpaque opaque;
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
OffsetNumber maxoffset;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
page = palloc(BLCKSZ);
/*
* We copy the page into local storage to avoid holding pin on the buffer
* longer than we must.
*/
buffer = ReadBufferExtended(state->rel, MAIN_FORKNUM, blocknum, RBM_NORMAL,
state->checkstrategy);
LockBuffer(buffer, BT_READ);
/*
* Perform the same basic sanity checking that nbtree itself performs for
* every page:
*/
_bt_checkpage(state->rel, buffer);
/* Only use copy of page in palloc()'d memory */
memcpy(page, BufferGetPage(buffer), BLCKSZ);
UnlockReleaseBuffer(buffer);
Add macros in hash and btree AMs to get the special area of their pages This makes the code more consistent with SpGiST, GiST and GIN, that already use this style, and the idea is to make easier the introduction of more sanity checks for each of these AM-specific macros. BRIN uses a different set of macros to get a page's type and flags, so it has no need for something similar. Author: Matthias van de Meent Discussion: https://postgr.es/m/CAEze2WjE3+tGO9Fs9+iZMU+z6mMZKo54W1Zt98WKqbEUHbHOBg@mail.gmail.com
2022-04-01 06:24:50 +02:00
opaque = BTPageGetOpaque(page);
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Minor code-cleanliness improvements for btree. Make the btree page-flags test macros (P_ISLEAF and friends) return clean boolean values, rather than values that might not fit in a bool. Use them in a few places that were randomly referencing the flag bits directly. In passing, change access/nbtree/'s only direct use of BUFFER_LOCK_SHARE to BT_READ. (Some think we should go the other way, but as long as we have BT_READ/BT_WRITE, let's use them consistently.) Masahiko Sawada, reviewed by Doug Doole Discussion: https://postgr.es/m/CAD21AoBmWPeN=WBB5Jvyz_Nt3rmW1ebUyAnk3ZbJP3RMXALJog@mail.gmail.com
2017-09-18 22:36:28 +02:00
if (P_ISMETA(opaque) && blocknum != BTREE_METAPAGE)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("invalid meta page found at block %u in index \"%s\"",
blocknum, RelationGetRelationName(state->rel))));
/* Check page from block that ought to be meta page */
if (blocknum == BTREE_METAPAGE)
{
BTMetaPageData *metad = BTPageGetMeta(page);
Minor code-cleanliness improvements for btree. Make the btree page-flags test macros (P_ISLEAF and friends) return clean boolean values, rather than values that might not fit in a bool. Use them in a few places that were randomly referencing the flag bits directly. In passing, change access/nbtree/'s only direct use of BUFFER_LOCK_SHARE to BT_READ. (Some think we should go the other way, but as long as we have BT_READ/BT_WRITE, let's use them consistently.) Masahiko Sawada, reviewed by Doug Doole Discussion: https://postgr.es/m/CAD21AoBmWPeN=WBB5Jvyz_Nt3rmW1ebUyAnk3ZbJP3RMXALJog@mail.gmail.com
2017-09-18 22:36:28 +02:00
if (!P_ISMETA(opaque) ||
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
metad->btm_magic != BTREE_MAGIC)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" meta page is corrupt",
RelationGetRelationName(state->rel))));
Skip full index scan during cleanup of B-tree indexes when possible Vacuum of index consists from two stages: multiple (zero of more) ambulkdelete calls and one amvacuumcleanup call. When workload on particular table is append-only, then autovacuum isn't intended to touch this table. However, user may run vacuum manually in order to fill visibility map and get benefits of index-only scans. Then ambulkdelete wouldn't be called for indexes of such table (because no heap tuples were deleted), only amvacuumcleanup would be called In this case, amvacuumcleanup would perform full index scan for two objectives: put recyclable pages into free space map and update index statistics. This patch allows btvacuumclanup to skip full index scan when two conditions are satisfied: no pages are going to be put into free space map and index statistics isn't stalled. In order to check first condition, we store oldest btpo_xact in the meta-page. When it's precedes RecentGlobalXmin, then there are some recyclable pages. In order to check second condition we store number of heap tuples observed during previous full index scan by cleanup. If fraction of newly inserted tuples is less than vacuum_cleanup_index_scale_factor, then statistics isn't considered to be stalled. vacuum_cleanup_index_scale_factor can be defined as both reloption and GUC (default). This patch bumps B-tree meta-page version. Upgrade of meta-page is performed "on the fly": during VACUUM meta-page is rewritten with new version. No special handling in pg_upgrade is required. Author: Masahiko Sawada, Alexander Korotkov Review by: Peter Geoghegan, Kyotaro Horiguchi, Alexander Korotkov, Yura Sokolov Discussion: https://www.postgresql.org/message-id/flat/CAD21AoAX+d2oD_nrd9O2YkpzHaFr=uQeGr9s1rKC3O4ENc568g@mail.gmail.com
2018-04-04 18:29:00 +02:00
if (metad->btm_version < BTREE_MIN_VERSION ||
metad->btm_version > BTREE_VERSION)
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
Skip full index scan during cleanup of B-tree indexes when possible Vacuum of index consists from two stages: multiple (zero of more) ambulkdelete calls and one amvacuumcleanup call. When workload on particular table is append-only, then autovacuum isn't intended to touch this table. However, user may run vacuum manually in order to fill visibility map and get benefits of index-only scans. Then ambulkdelete wouldn't be called for indexes of such table (because no heap tuples were deleted), only amvacuumcleanup would be called In this case, amvacuumcleanup would perform full index scan for two objectives: put recyclable pages into free space map and update index statistics. This patch allows btvacuumclanup to skip full index scan when two conditions are satisfied: no pages are going to be put into free space map and index statistics isn't stalled. In order to check first condition, we store oldest btpo_xact in the meta-page. When it's precedes RecentGlobalXmin, then there are some recyclable pages. In order to check second condition we store number of heap tuples observed during previous full index scan by cleanup. If fraction of newly inserted tuples is less than vacuum_cleanup_index_scale_factor, then statistics isn't considered to be stalled. vacuum_cleanup_index_scale_factor can be defined as both reloption and GUC (default). This patch bumps B-tree meta-page version. Upgrade of meta-page is performed "on the fly": during VACUUM meta-page is rewritten with new version. No special handling in pg_upgrade is required. Author: Masahiko Sawada, Alexander Korotkov Review by: Peter Geoghegan, Kyotaro Horiguchi, Alexander Korotkov, Yura Sokolov Discussion: https://www.postgresql.org/message-id/flat/CAD21AoAX+d2oD_nrd9O2YkpzHaFr=uQeGr9s1rKC3O4ENc568g@mail.gmail.com
2018-04-04 18:29:00 +02:00
errmsg("version mismatch in index \"%s\": file version %d, "
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
"current version %d, minimum supported version %d",
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
RelationGetRelationName(state->rel),
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
metad->btm_version, BTREE_VERSION,
BTREE_MIN_VERSION)));
/* Finished with metapage checks */
return page;
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
}
/*
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
* Deleted pages that still use the old 32-bit XID representation have no
* sane "level" field because they type pun the field, but all other pages
* (including pages deleted on Postgres 14+) have a valid value.
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
*/
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
if (!P_ISDELETED(opaque) || P_HAS_FULLXID(opaque))
{
/* Okay, no reason not to trust btpo_level field from page */
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
if (P_ISLEAF(opaque) && opaque->btpo_level != 0)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("invalid leaf page level %u for block %u in index \"%s\"",
opaque->btpo_level, blocknum,
RelationGetRelationName(state->rel))));
if (!P_ISLEAF(opaque) && opaque->btpo_level == 0)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("invalid internal page level 0 for block %u in index \"%s\"",
blocknum,
RelationGetRelationName(state->rel))));
}
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
/*
* Sanity checks for number of items on page.
*
* As noted at the beginning of _bt_binsrch(), an internal page must have
* children, since there must always be a negative infinity downlink
* (there may also be a highkey). In the case of non-rightmost leaf
Remove btree page items after page unlink Currently, page unlink leaves remaining items "as is", but replay of corresponding WAL-record re-initializes page leaving it with no items. For the sake of consistency, this commit makes primary delete all the items during page unlink as well. Thanks to this change, we now don't mask contents of deleted btree page for WAL consistency checking. Discussion: https://postgr.es/m/CAPpHfdt_OTyQpXaPJcWzV2N-LNeNJseNB-K_A66qG%3DL518VTFw%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan
2020-08-05 01:15:34 +02:00
* pages, there must be at least a highkey. The exceptions are deleted
* pages, which contain no items.
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
*
* This is correct when pages are half-dead, since internal pages are
* never half-dead, and leaf pages must have a high key when half-dead
* (the rightmost page can never be deleted). It's also correct with
* fully deleted pages: _bt_unlink_halfdead_page() doesn't change anything
* about the target page other than setting the page as fully dead, and
* setting its xact field. In particular, it doesn't change the sibling
* links in the deletion target itself, since they're required when index
* scans land on the deletion target, and then need to move right (or need
* to move left, in the case of backward index scans).
*/
maxoffset = PageGetMaxOffsetNumber(page);
if (maxoffset > MaxIndexTuplesPerPage)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("Number of items on block %u of index \"%s\" exceeds MaxIndexTuplesPerPage (%u)",
blocknum, RelationGetRelationName(state->rel),
MaxIndexTuplesPerPage)));
Fix amcheck for page checks concurrent to replay of btree page deletion amcheck expects at least hikey to always exist on leaf page even if it is deleted page. But replica reinitializes page during replay of page deletion, causing deleted page to have no items. Thus, replay of page deletion can cause an error in concurrent amcheck run. This commit relaxes amcheck expectation making it tolerate deleted page with no items. Reported-by: Konstantin Knizhnik Discussion: https://postgr.es/m/CAPpHfdt_OTyQpXaPJcWzV2N-LNeNJseNB-K_A66qG%3DL518VTFw%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan Backpatch-through: 11
2020-05-14 11:44:44 +02:00
if (!P_ISLEAF(opaque) && !P_ISDELETED(opaque) && maxoffset < P_FIRSTDATAKEY(opaque))
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("internal block %u in index \"%s\" lacks high key and/or at least one downlink",
blocknum, RelationGetRelationName(state->rel))));
Fix amcheck for page checks concurrent to replay of btree page deletion amcheck expects at least hikey to always exist on leaf page even if it is deleted page. But replica reinitializes page during replay of page deletion, causing deleted page to have no items. Thus, replay of page deletion can cause an error in concurrent amcheck run. This commit relaxes amcheck expectation making it tolerate deleted page with no items. Reported-by: Konstantin Knizhnik Discussion: https://postgr.es/m/CAPpHfdt_OTyQpXaPJcWzV2N-LNeNJseNB-K_A66qG%3DL518VTFw%40mail.gmail.com Author: Alexander Korotkov Reviewed-by: Peter Geoghegan Backpatch-through: 11
2020-05-14 11:44:44 +02:00
if (P_ISLEAF(opaque) && !P_ISDELETED(opaque) && !P_RIGHTMOST(opaque) && maxoffset < P_HIKEY)
Add missing and dangling downlink checks to amcheck When bt_index_parent_check() is called with the heapallindexed option, allocate a second Bloom filter to fingerprint block numbers that appear in the downlinks of internal pages. Use Bloom filter probes when walking the B-Tree to detect missing downlinks. This can detect subtle problems with page deletion/VACUUM, such as corruption caused by the bug just fixed in commit 6db4b499. The downlink Bloom filter is bound in size by work_mem. Its optimal size is typically far smaller than that of the regular heapallindexed Bloom filter, especially when the index has high fan-out. Author: Peter Geoghegan Reviewer: Teodor Sigaev Discussion: https://postgr.es/m/CAH2-WznUzY4fWTjm1tBB3JpVz8cCfz7k_qVp5BhuPyhivmWJFg@mail.gmail.com
2018-04-25 17:02:55 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("non-rightmost leaf block %u in index \"%s\" lacks high key item",
blocknum, RelationGetRelationName(state->rel))));
/*
* In general, internal pages are never marked half-dead, except on
* versions of Postgres prior to 9.4, where it can be valid transient
* state. This state is nonetheless treated as corruption by VACUUM on
* from version 9.4 on, so do the same here. See _bt_pagedel() for full
* details.
*/
if (!P_ISLEAF(opaque) && P_ISHALFDEAD(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("internal page block %u in index \"%s\" is half-dead",
blocknum, RelationGetRelationName(state->rel)),
errhint("This can be caused by an interrupted VACUUM in version 9.3 or older, before upgrade. Please REINDEX it.")));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
/*
* Check that internal pages have no garbage items, and that no page has
* an invalid combination of deletion-related page level flags
*/
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
if (!P_ISLEAF(opaque) && P_HAS_GARBAGE(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
Use full 64-bit XIDs in deleted nbtree pages. Otherwise we risk "leaking" deleted pages by making them non-recyclable indefinitely. Commit 6655a729 did the same thing for deleted pages in GiST indexes. That work was used as a starting point here. Stop storing an XID indicating the oldest bpto.xact across all deleted though unrecycled pages in nbtree metapages. There is no longer any reason to care about that condition/the oldest XID. It only ever made sense when wraparound was something _bt_vacuum_needs_cleanup() had to consider. The btm_oldest_btpo_xact metapage field has been repurposed and renamed. It is now btm_last_cleanup_num_delpages, which is used to remember how many non-recycled deleted pages remain from the last VACUUM (in practice its value is usually the precise number of pages that were _newly deleted_ during the specific VACUUM operation that last set the field). The general idea behind storing btm_last_cleanup_num_delpages is to use it to give _some_ consideration to non-recycled deleted pages inside _bt_vacuum_needs_cleanup() -- though never too much. We only really need to avoid leaving a truly excessive number of deleted pages in an unrecycled state forever. We only do this to cover certain narrow cases where no other factor makes VACUUM do a full scan, and yet the index continues to grow (and so actually misses out on recycling existing deleted pages). These metapage changes result in a clear user-visible benefit: We no longer trigger full index scans during VACUUM operations solely due to the presence of only 1 or 2 known deleted (though unrecycled) blocks from a very large index. All that matters now is keeping the costs and benefits in balance over time. Fix an issue that has been around since commit 857f9c36, which added the "skip full scan of index" mechanism (i.e. the _bt_vacuum_needs_cleanup() logic). The accuracy of btm_last_cleanup_num_heap_tuples accidentally hinged upon _when_ the source value gets stored. We now always store btm_last_cleanup_num_heap_tuples in btvacuumcleanup(). This fixes the issue because IndexVacuumInfo.num_heap_tuples (the source field) is expected to accurately indicate the state of the table _after_ the VACUUM completes inside btvacuumcleanup(). A backpatchable fix cannot easily be extracted from this commit. A targeted fix for the issue will follow in a later commit, though that won't happen today. I (pgeoghegan) have chosen to remove any mention of deleted pages in the documentation of the vacuum_cleanup_index_scale_factor GUC/param, since the presence of deleted (though unrecycled) pages is no longer of much concern to users. The vacuum_cleanup_index_scale_factor description in the docs now seems rather unclear in any case, and it should probably be rewritten in the near future. Perhaps some passing mention of page deletion will be added back at the same time. Bump XLOG_PAGE_MAGIC due to nbtree WAL records using full XIDs now. Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/CAH2-WznpdHvujGUwYZ8sihX=d5u-tRYhi-F4wnV2uN2zHpMUXw@mail.gmail.com
2021-02-25 03:41:34 +01:00
errmsg_internal("internal page block %u in index \"%s\" has garbage items",
blocknum, RelationGetRelationName(state->rel))));
if (P_HAS_FULLXID(opaque) && !P_ISDELETED(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("full transaction id page flag appears in non-deleted block %u in index \"%s\"",
blocknum, RelationGetRelationName(state->rel))));
if (P_ISDELETED(opaque) && P_ISHALFDEAD(opaque))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("deleted page block %u in index \"%s\" is half-dead",
blocknum, RelationGetRelationName(state->rel))));
Add amcheck extension to contrib. This is the beginning of a collection of SQL-callable functions to verify the integrity of data files. For now it only contains code to verify B-Tree indexes. This adds two SQL-callable functions, validating B-Tree consistency to a varying degree. Check the, extensive, docs for details. The goal is to later extend the coverage of the module to further access methods, possibly including the heap. Once checks for additional access methods exist, we'll likely add some "dispatch" functions that cover multiple access methods. Author: Peter Geoghegan, editorialized by Andres Freund Reviewed-By: Andres Freund, Tomas Vondra, Thomas Munro, Anastasia Lubennikova, Robert Haas, Amit Langote Discussion: CAM3SWZQzLMhMwmBqjzK+pRKXrNUZ4w90wYMUWfkeV8mZ3Debvw@mail.gmail.com
2017-03-10 00:50:40 +01:00
return page;
}
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/*
* _bt_mkscankey() wrapper that automatically prevents insertion scankey from
* being considered greater than the pivot tuple that its values originated
* from (or some other identical pivot tuple) in the common case where there
* are truncated/minus infinity attributes. Without this extra step, there
* are forms of corruption that amcheck could theoretically fail to report.
*
* For example, invariant_g_offset() might miss a cross-page invariant failure
* on an internal level if the scankey built from the first item on the
* target's right sibling page happened to be equal to (not greater than) the
* last item on target page. The !pivotsearch tiebreaker in _bt_compare()
* might otherwise cause amcheck to assume (rather than actually verify) that
* the scankey is greater.
*/
static inline BTScanInsert
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
bt_mkscankey_pivotsearch(Relation rel, Relation heaprel, IndexTuple itup)
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
{
BTScanInsert skey;
Pass down table relation into more index relation functions This is done in preparation for logical decoding on standby, which needs to include whether visibility affecting WAL records are about a (user) catalog table. Which is only known for the table, not the indexes. It's also nice to be able to pass the heap relation to GlobalVisTestFor() in vacuumRedirectAndPlaceholder(). Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com> Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
2023-04-02 05:12:26 +02:00
skey = _bt_mkscankey(rel, heaprel, itup);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
skey->pivotsearch = true;
return skey;
}
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
/*
* PageGetItemId() wrapper that validates returned line pointer.
*
* Buffer page/page item access macros generally trust that line pointers are
* not corrupt, which might cause problems for verification itself. For
* example, there is no bounds checking in PageGetItem(). Passing it a
* corrupt line pointer can cause it to return a tuple/pointer that is unsafe
* to dereference.
*
* Validating line pointers before tuples avoids undefined behavior and
* assertion failures with corrupt indexes, making the verification process
* more robust and predictable.
*/
static ItemId
PageGetItemIdCareful(BtreeCheckState *state, BlockNumber block, Page page,
OffsetNumber offset)
{
ItemId itemid = PageGetItemId(page, offset);
if (ItemIdGetOffset(itemid) + ItemIdGetLength(itemid) >
amcheck: MAXALIGN() nbtree special area offset. This isn't strictly necessary, but in theory it might matter if in the future the width of the nbtree special area changes -- its total size might not be an even number of MAXALIGN() quantums, even with padding. PageInit() MAXALIGN()s all special area offsets, but amcheck uses the offset to perform initial basic validation of line pointers, so we don't rely on the offset from the page header. The real reason to do this is to set a good example for new code that adds amcheck coverage for other index AMs. Reported-By: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com> Discussion: https://postgr.es/m/CALj2ACUMqTR9nErh99FbOBmzCXE9=gXNqhBiwYOhejJJS1LXqQ@mail.gmail.com
2021-04-24 00:37:03 +02:00
BLCKSZ - MAXALIGN(sizeof(BTPageOpaqueData)))
Sanitize line pointers within contrib/amcheck. Adopt a more defensive approach to accessing index tuples in contrib/amcheck: verify that each line pointer looks sane before accessing associated tuple using pointer arithmetic based on line pointer's offset. This avoids undefined behavior and assertion failures in cases where line pointers are corrupt. Issue spotted following a complaint about an assertion failure by Grigory Smolkin, which involved a test harness that deliberately corrupts indexes. This is arguably a bugfix, but no backpatch given the lack of field reports beyond Grigory's. Discussion: https://postgr.es/m/CAH2-WzmkurhCqnyLHxk0VkOZqd49+ZZsp1xAJOg7j2x7dmp_XQ@mail.gmail.com
2019-04-25 21:50:37 +02:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("line pointer points past end of tuple space in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=(%u,%u) lp_off=%u, lp_len=%u lp_flags=%u.",
block, offset, ItemIdGetOffset(itemid),
ItemIdGetLength(itemid),
ItemIdGetFlags(itemid))));
/*
* Verify that line pointer isn't LP_REDIRECT or LP_UNUSED, since nbtree
* never uses either. Verify that line pointer has storage, too, since
* even LP_DEAD items should within nbtree.
*/
if (ItemIdIsRedirected(itemid) || !ItemIdIsUsed(itemid) ||
ItemIdGetLength(itemid) == 0)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("invalid line pointer storage in index \"%s\"",
RelationGetRelationName(state->rel)),
errdetail_internal("Index tid=(%u,%u) lp_off=%u, lp_len=%u lp_flags=%u.",
block, offset, ItemIdGetOffset(itemid),
ItemIdGetLength(itemid),
ItemIdGetFlags(itemid))));
return itemid;
}
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
/*
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
* BTreeTupleGetHeapTID() wrapper that enforces that a heap TID is present in
* cases where that is mandatory (i.e. for non-pivot tuples)
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
*/
static inline ItemPointer
BTreeTupleGetHeapTIDCareful(BtreeCheckState *state, IndexTuple itup,
bool nonpivot)
{
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
ItemPointer htid;
/*
* Caller determines whether this is supposed to be a pivot or non-pivot
* tuple using page type and item offset number. Verify that tuple
* metadata agrees with this.
*/
Assert(state->heapkeyspace);
if (BTreeTupleIsPivot(itup) && nonpivot)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("block %u or its right sibling block or child block in index \"%s\" has unexpected pivot tuple",
state->targetblock,
RelationGetRelationName(state->rel))));
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
if (!BTreeTupleIsPivot(itup) && !nonpivot)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg_internal("block %u or its right sibling block or child block in index \"%s\" has unexpected non-pivot tuple",
state->targetblock,
RelationGetRelationName(state->rel))));
htid = BTreeTupleGetHeapTID(itup);
if (!ItemPointerIsValid(htid) && nonpivot)
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("block %u or its right sibling block or child block in index \"%s\" contains non-pivot tuple that lacks a heap TID",
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
state->targetblock,
RelationGetRelationName(state->rel))));
return htid;
}
/*
* Return the "pointed to" TID for itup, which is used to generate a
* descriptive error message. itup must be a "data item" tuple (it wouldn't
* make much sense to call here with a high key tuple, since there won't be a
* valid downlink/block number to display).
*
* Returns either a heap TID (which will be the first heap TID in posting list
* if itup is posting list tuple), or a TID that contains downlink block
* number, plus some encoded metadata (e.g., the number of attributes present
* in itup).
*/
static inline ItemPointer
BTreeTupleGetPointsToTID(IndexTuple itup)
{
/*
* Rely on the assumption that !heapkeyspace internal page data items will
* correctly return TID with downlink here -- BTreeTupleGetHeapTID() won't
* recognize it as a pivot tuple, but everything still works out because
* the t_tid field is still returned
*/
if (!BTreeTupleIsPivot(itup))
return BTreeTupleGetHeapTID(itup);
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
Add deduplication to nbtree. Deduplication reduces the storage overhead of duplicates in indexes that use the standard nbtree index access method. The deduplication process is applied lazily, after the point where opportunistic deletion of LP_DEAD-marked index tuples occurs. Deduplication is only applied at the point where a leaf page split would otherwise be required. New posting list tuples are formed by merging together existing duplicate tuples. The physical representation of the items on an nbtree leaf page is made more space efficient by deduplication, but the logical contents of the page are not changed. Even unique indexes make use of deduplication as a way of controlling bloat from duplicates whose TIDs point to different versions of the same logical table row. The lazy approach taken by nbtree has significant advantages over a GIN style eager approach. Most individual inserts of index tuples have exactly the same overhead as before. The extra overhead of deduplication is amortized across insertions, just like the overhead of page splits. The key space of indexes works in the same way as it has since commit dd299df8 (the commit that made heap TID a tiebreaker column). Testing has shown that nbtree deduplication can generally make indexes with about 10 or 15 tuples for each distinct key value about 2.5X - 4X smaller, even with single column integer indexes (e.g., an index on a referencing column that accompanies a foreign key). The final size of single column nbtree indexes comes close to the final size of a similar contrib/btree_gin index, at least in cases where GIN's posting list compression isn't very effective. This can significantly improve transaction throughput, and significantly reduce the cost of vacuuming indexes. A new index storage parameter (deduplicate_items) controls the use of deduplication. The default setting is 'on', so all new B-Tree indexes automatically use deduplication where possible. This decision will be reviewed at the end of the Postgres 13 beta period. There is a regression of approximately 2% of transaction throughput with synthetic workloads that consist of append-only inserts into a table with several non-unique indexes, where all indexes have few or no repeated values. The underlying issue is that cycles are wasted on unsuccessful attempts at deduplicating items in non-unique indexes. There doesn't seem to be a way around it short of disabling deduplication entirely. Note that deduplication of items in unique indexes is fairly well targeted in general, which avoids the problem there (we can use a special heuristic to trigger deduplication passes in unique indexes, since we're specifically targeting "version bloat"). Bump XLOG_PAGE_MAGIC because xl_btree_vacuum changed. No bump in BTREE_VERSION, since the representation of posting list tuples works in a way that's backwards compatible with version 4 indexes (i.e. indexes built on PostgreSQL 12). However, users must still REINDEX a pg_upgrade'd index to use deduplication, regardless of the Postgres version they've upgraded from. This is the only way to set the new nbtree metapage flag indicating that deduplication is generally safe. Author: Anastasia Lubennikova, Peter Geoghegan Reviewed-By: Peter Geoghegan, Heikki Linnakangas Discussion: https://postgr.es/m/55E4051B.7020209@postgrespro.ru https://postgr.es/m/4ab6e2db-bcee-f4cf-0916-3a06e6ccbb55@postgrespro.ru
2020-02-26 22:05:30 +01:00
/* Pivot tuple returns TID with downlink block (heapkeyspace variant) */
return &itup->t_tid;
Make heap TID a tiebreaker nbtree index column. Make nbtree treat all index tuples as having a heap TID attribute. Index searches can distinguish duplicates by heap TID, since heap TID is always guaranteed to be unique. This general approach has numerous benefits for performance, and is prerequisite to teaching VACUUM to perform "retail index tuple deletion". Naively adding a new attribute to every pivot tuple has unacceptable overhead (it bloats internal pages), so suffix truncation of pivot tuples is added. This will usually truncate away the "extra" heap TID attribute from pivot tuples during a leaf page split, and may also truncate away additional user attributes. This can increase fan-out, especially in a multi-column index. Truncation can only occur at the attribute granularity, which isn't particularly effective, but works well enough for now. A future patch may add support for truncating "within" text attributes by generating truncated key values using new opclass infrastructure. Only new indexes (BTREE_VERSION 4 indexes) will have insertions that treat heap TID as a tiebreaker attribute, or will have pivot tuples undergo suffix truncation during a leaf page split (on-disk compatibility with versions 2 and 3 is preserved). Upgrades to version 4 cannot be performed on-the-fly, unlike upgrades from version 2 to version 3. contrib/amcheck continues to work with version 2 and 3 indexes, while also enforcing stricter invariants when verifying version 4 indexes. These stricter invariants are the same invariants described by "3.1.12 Sequencing" from the Lehman and Yao paper. A later patch will enhance the logic used by nbtree to pick a split point. This patch is likely to negatively impact performance without smarter choices around the precise point to split leaf pages at. Making these two mostly-distinct sets of enhancements into distinct commits seems like it might clarify their design, even though neither commit is particularly useful on its own. The maximum allowed size of new tuples is reduced by an amount equal to the space required to store an extra MAXALIGN()'d TID in a new high key during leaf page splits. The user-facing definition of the "1/3 of a page" restriction is already imprecise, and so does not need to be revised. However, there should be a compatibility note in the v12 release notes. Author: Peter Geoghegan Reviewed-By: Heikki Linnakangas, Alexander Korotkov Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
2019-03-20 18:04:01 +01:00
}