Jul - Aug 2022

5 commits
Improve recently-added test reliability
6 Aug 2022Alvaro Herreramaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15beta3REL_14_STABLE14.5REL_13_STABLE13.8+3 more
LLM analysis, 100% surewritten from the discussion, no matching patch3 of 3 reviewers saw the final version

Implements Thomas's diagnosis in #130 by configuring primary_slot_name and retaining the slot, with catch-up and polling cleanups in 033_replay_tsp_drops.pl. No attachment contains these changes; Alvaro promises this repair in #134.

  • Thomas Munrosaw the final versionLLM analysis, 99% sureDiagnosed the slot-drop race in the committed version of #124 and proposed configuring the standby to use the slot in #130.
  • Tom Lanesaw the final versionLLM analysis, 97% sureReported that buildfarm failures continued after the first slot fix in #126, prompting the additional reliability repair.
  • Andres Freundsaw the final versionLLM analysis, 84% sureCorroborated the remaining test failure with two additional CI reproductions in #131.
Fix test instability
29 Jul 2022Alvaro Herreramaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15beta3REL_14_STABLE14.5REL_13_STABLE13.8+3 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #1244 of 4 reviewers saw the final version

Uses the replication-slot fix from #124, moving and renaming the slot creation, adding a slot drop after connection detection, and correcting a comment typo. Alvaro explicitly describes the added slot drop in #125.

  • Matthias van de Meentsaw the final versionLLM analysis, 99% sureReported intermittent standby catch-up failures in the newly committed test in #121, leading to the slot fix.
  • Tom Lanesaw the final versionLLM analysis, 97% sureConfirmed the failure on the buildfarm and identified a likely timing dependency in #122.
  • Thomas Munrosaw the final versionLLM analysis, 100% sureIdentified premature WAL removal and proposed a replication slot in #123, which the author implemented in #124.
  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Fix replay of create database records on standby
28 Jul 2022Alvaro Herreramaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15beta3REL_14_STABLE14.5REL_13_STABLE13.8+3 more
LLM analysis, 100% surelanded as postedfrom patchset #1201 of 4 reviewers saw the final version

All four file diffs match #120, including recovery_create_dbdir(), CheckTablespaceDirectory(), and the 169-line TAP test. The forwarded patch credits Kyotaro, Asim, and Paul; Kyotaro supplied the replacement implementation in #96.

  • Anastasia Lubennikovasaw an earlier versiontrailerFound unsafe archive-recovery behavior in #23 and tested the hash-based revision in #32, before the later replacement of that design.
  • Fujii Masaosaw an earlier versiontrailerIdentified the minimum-recovery-point problem and diagnostic issues in #37, well before the directory-reconstruction rewrite.
  • MichaĆ«l Paquiersaw an earlier versiontrailerReviewed test infrastructure and the hash-based fix, proposing in-place tablespaces and standalone tests in #68, with no review of the later reconstruction implementation shown.
  • Alvaro Herrerasaw the final versioncommittercommitted the patch
Allow "in place" tablespaces.
27 Jul 2022Alvaro HerreraREL_14_STABLE14.5REL_13_STABLE13.8REL_12_STABLE12.12REL_11_STABLE11.17REL_10_STABLE10.22
LLM analysis, 100% surewritten from the discussion, no matching patch0 of 1 reviewer saw the final version

This supplied candidate is the stable-branch rollup of in-place tablespace support and related fixes, assembled to enable the recovery backpatch rather than matching an attachment. #119 identifies the bundle; the commit credits Thomas, Michael, and Alvaro.

  • Kyotaro Horiguchisaw an earlier versionLLM analysis, 92% sureSupported backpatching the developer feature in #114 and examined its dependencies in #117, before Alvaro expanded the bundle in #119.
Fix get_dirent_type() for Windows junction points.
22 Jul 2022Thomas Munromaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1REL_15_STABLE15beta3REL_14_STABLE14.5Also in: pg_tablespace_location() failure with allow_in_place_tablespaces (part of a posted patchset)

Mar 2022

2 commits
Revert "Fix replay of create database records on standby"
29 Mar 2022Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7reverts 1 commit here
LLM analysis, 100% surewritten from the discussion, no matching patch0 of 1 reviewer saw the final version

Reverses 49d9cfc68b following Robert's inconsistent-recovery-strategy diagnosis in #82 and Alvaro's explicit revert decision in #86. No revert patch was attached.

  • Robert Haassaw an earlier versionLLM analysis, 100% sureExplained why skipped database copies conflicted with existing directory-creating redo in #82 and accepted the replacement approach in #87, before the later major rework.
Fix replay of create database records on standby
25 Mar 2022Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14.3REL_13_STABLE13.7
LLM analysis, 100% surelanded as postedfrom patchset #741 of 5 reviewers saw the final version

All seven file diffs match #74, including missing_dir_tab, the pre-consistency flush guard, typedef entries, and the TAP test. Alvaro forwarded and committed work credited to Paul, Kyotaro, and Asim.

  • Anastasia Lubennikovasaw an earlier versionLLM analysis, 99% sureReported the archive-recovery correctness bug in #23 and tested the rebased hash implementation in #32, before further recovery-logic changes.
  • Fujii Masaosaw an earlier versionLLM analysis, 100% sureProposed advancing minimum recovery before tablespace deletion and corrected diagnostic handling in #37, with those logic changes incorporated later.
  • Tom Lanesaw an earlier versionLLM analysis, 96% sureAssessed the backend approach as sane and requested removal of unused polling changes in #50, before the final flush restriction and test redesign.
  • Michael Paquiersaw an earlier versionLLM analysis, 97% sureProposed the adopted in-place-tablespace test simplification and examined the backend paths in #68, before the XLogFlush condition changed in #74.
  • Alvaro Herrerasaw the final versioncommittercommitted the patch

Credits across 7 commits

Author
Kyotaro Horiguchi3
Asim R P2
Paul Guo2
Thomas Munro1 LLM2
Alvaro Herrera1
Michael Paquier1
Reviewer
Alvaro Herreracommitter4
Michael Paquier2
Anastasia Lubennikova1
Fujii Masao1
Reported by
Matthias van de Meent1
Committer
Alvaro Herrera6
Thomas Munro1

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