Mar - Apr 2023

11 commits
Add unit to vacuum_buffer_usage_limit value in postgresql.conf.sample.
28 Apr 2023Masahiko Sawadamaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surelanded as postedfrom patchset #753 of 3 reviewers saw the final version

The added and removed lines in postgresql.conf.sample exactly match #75, including the 256kB value and comment alignment.

  • David Rowleysaw the final versiontrailerIn #76, he endorsed adding the unit consistently with work_mem and supported committing the adjustment.
  • Daniel Gustafssonsaw the final versiontrailerIn #77, he confirmed related sample configuration entries use units and endorsed the patch.
  • Melanie Plagemansaw the final versiontrailerIn #78, she approved the patch while checking tab alignment before the comment.
Improve VACUUM/ANALYZE BUFFER_USAGE_LIMIT docs
16 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #711 of 1 reviewer saw the final version

Matches the four-file documentation/help correction in #71, with only wording changes specifying a string containing the numerical size in the VACUUM and ANALYZE manuals.

  • Peter Eisentrautsaw the final versionLLM analysis, 93% sureHis review of the landed SQL and vacuumdb documentation in #70 identified the inconsistent placeholders, unexplained units, and option ordering corrected by #71.
Add --buffer-usage-limit option to vacuumdb
7 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #631 of 2 reviewers saw the final version

Takes the vacuumdb component of #63, permitting --full when --analyze is also present. It additionally corrects KB to kB in server errors, configuration documentation, and expected test output, as explained in #69.

  • Justin Pryzbysaw an earlier versiontrailerHe requested vacuumdb support in #20, reviewed its documentation in #30, and discussed FULL compatibility in #43, before later option-handling changes.
  • David Rowleysaw the final versiontrailerHe reviewed and committed the vacuumdb component, describing the final compatibility and unit-message adjustments in #69.
Add VACUUM/ANALYZE BUFFER_USAGE_LIMIT option
6 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #662 of 4 reviewers saw the final version

The main feature matches David Rowley's repost of Melanie's work in #66, with expanded performance guidance, README corrections, an equivalent simplification of default-size selection, and an additional ANALYZE zero-limit test.

  • David Rowleysaw the final versiontrailerHe performed detailed reviews in #48 and #50, supplied the final revisions, and reported committing them in #69.
  • Andres Freundsaw an earlier versiontrailerHis reviews in #15, #52, and #55 shaped the size API and allocation refactoring, before later default semantics and parallel NULL handling changed.
  • Justin Pryzbysaw the final versiontrailerAfter earlier interface and documentation reviews, he caught the final README wording defect in #67, which the committer explicitly incorporated.
  • Bharath Rupireddysaw an earlier versiontrailerHe reviewed validation, error messages, and the strategy API in #17, before the size-based GUC interface and subsequent allocation refactoring.
Move various prechecks from vacuum() into ExecVacuum()
6 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #631 of 1 reviewer saw the final version

Commits the setup-refactoring component of #63, retaining transaction-block detection inside vacuum() rather than passing in_outer_xact from ExecVacuum(). It also creates the autovacuum context under CurrentMemoryContext instead of adding a portal-context argument.

  • David Rowleysaw the final versiontrailerHe examined strategy lifetime in #60, corrected his concern in #62, and reviewed the final implementation as its committer.
Always make a BufferAccessStrategy for ANALYZE
6 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

David Rowley's correction follows Melanie's FULL/ANALYZE regression report and inline proposal in #58. It was committed at 00:37 UTC, before the corrected standalone attachment in #59 at 00:41 UTC, and uses a different condition.

  • Melanie Plagemansaw the final versionLLM analysis, 97% sureIn #58, she identified that the committed allocation optimization incorrectly excluded VACUUM FULL with ANALYZE, directly prompting this correction.
Rename BufferAccessStrategyData.ring_size to nbuffers
3 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #441 of 1 reviewer saw the final version

Exactly matches the ring_size-to-nbuffers component of #44 in src/backend/storage/buffer/freelist.c; the other series components were committed separately.

  • David Rowleysaw the final versiontrailerHe committed the rename component and explicitly confirmed that landing in #46.
Disable vacuum's use of a buffer access strategy during failsafe
3 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #442 of 2 reviewers saw the final version

Commits the failsafe component of #44, preserving vacrel->bstrategy = NULL while revising comments and documentation. It also includes an equivalent simplification and explanatory comment for the preceding allocation condition in vacuum.c.

  • Justin Pryzbysaw the final versiontrailerHe requested a clearer explanation of the failsafe bypass in #20, after which this component received only explanatory and equivalent-expression changes.
  • David Rowleysaw the final versiontrailerAs committer, he reviewed the final failsafe change and supplied its documentation and allocation-comment adjustments.
Only make buffer strategy for vacuum when it's likely needed
3 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #441 of 1 reviewer saw the final version

The vacuum.c diff exactly matches the conditional strategy-allocation component of #44, excluding the other components of that series.

  • David Rowleysaw the final versiontrailerHe reviewed the final allocation optimization as its committer, although the thread contains no separate detailed review of this component.
Remove some global variables from vacuum.c
3 Apr 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 99% suremodified before commitlight changesfrom patchset #391 of 2 reviewers saw the final version

Matches David Rowley's expanded replacement in #39, removing both vacuum.c globals, with only expand_vacuum_rel argument ordering changed. The commit credits Melanie Plageman, whose original globals-removal patch David extended.

  • Bharath Rupireddysaw an earlier versiontrailerHe reviewed the original vac_strategy removal in #17, before the replacement also removed vac_context and changed relation-list helper interfaces.
  • David Rowleysaw the final versioncommittercommitted the patch
Doc: add Buffer Access Strategy to the glossary
31 Mar 2023David Rowleymaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 99% suremodified before commitlight changesfrom patchset #362 of 2 reviewers saw the final version

Combines the glossary addition in #31 with the successive fixups #35 and #36. Final edits change associated WAL to unflushed WAL, remove CTAS, and adjust wrapping, as explained in #37.

  • Justin Pryzbysaw the final versiontrailerIn #30, he supplied glossary wording that Melanie incorporated into #31, followed by documentation-only refinements.
  • David Rowleysaw the final versiontrailerHe supplied glossary markup and wording fixes in #35, reviewed the follow-up, and explained his final edits and commit in #37.

Credits across 11 commits

Author
Melanie Plageman8
David RowleyLLM1
Masahiko Sawada1
Reviewer
David Rowleycommitter9
Justin Pryzby4
Bharath Rupireddy2
Andres Freund1
Daniel Gustafsson1
Melanie Plageman1
Reported by
Melanie Plageman1
Peter Eisentraut1
Committer
David Rowley10
Masahiko Sawada1

N LLM means N of the credits come from LLM analysis, not from commit trailers.