Log In - Home Page

Patch: REINDEX CONCURRENTLY

Edit Patch - Move To Another CommitFest - Delete Patch

CommitFest 2013-11
Topic Server Features
Patch Status Returned with Feedback
Author Michael Paquier
Reviewers Nobody
Committer Nobody
Close Date 2013-11-27
Comments
Patch by michael-kun on 2012-10-03 05:32:25 AM: Initial version.
Patch by michael-kun on 2012-10-13 02:59:59 AM: v2: support for toast, correct behavior of constraint indexes, correct drop + bug fixes
Patch by michael-kun on 2012-12-08 12:36:56 PM: v3: Realignment with head of 6th Dec 2012. Fix for old index drop now really concurrent. Swap is now 1 TX/index.
Patch by michael-kun on 2012-12-10 12:21:07 PM: v4: update reltoastidxid for toast parent relation at swap phase
Review by andresfreund on 2012-12-10 03:41:11 PM: Review of v4
Patch by michael-kun on 2012-12-17 02:57:44 AM: v5: Fix for fkey reference. Refactor APIs. Comments of v4 review.
Review by michael-kun on 2013-01-16 06:37:03 AM: Review of v5 by Andres
Patch by michael-kun on 2013-01-16 06:37:54 AM: v6: correction of comments after review of v5
Patch by michael-kun on 2013-01-16 06:38:33 AM: v7: support for REINDEX DATABASE CONCURRENTLY
Review by andresfreund on 2013-01-23 07:09:12 PM: Review for v7, good progress, some adjustments needed, waiting for reply from author
Comment by ringerc on 2013-01-25 03:41:43 AM: Just a macro/helper for things you can do already? (Tom)
Comment by ringerc on 2013-01-25 03:42:30 AM: Doing via script has significant limitations resolved by REINDEX CONCURRENTLY, + usability issues. (Andres)
Patch by michael-kun on 2013-01-25 05:15:32 AM: v8: corrected after review of v7
Patch by michael-kun on 2013-02-14 07:11:40 AM: v11: 3 patches: 1) removal of reltoastidxid, 2) feature, 3) docs. Concurrent build of toast indexes is fully supported thanks to patch 1. Contains all previous fixes of past reviews.
Review by michael-kun on 2013-02-21 02:59:57 AM: Review by Fujii Masao. Deadlock detected when REINDEX CONCURRENTLY is done in parallel with unexpected concurrent indexes.
Patch by michael-kun on 2013-02-21 03:00:39 AM: v12: fix deadlock problem by disallowing parallel REINDEX CONCURRENTLY to run in parallel for the same relations
Review by masaofujii on 2013-02-22 05:17:26 PM: Deadlock happened again.
Comment by michael-kun on 2013-02-25 03:01:58 AM: Deadlock risks also exist with CREATE INDEX CONCURRENTLY, in this case indexes are marked as invalid when a deadlock occurs.
Patch by michael-kun on 2013-02-25 03:03:07 AM: v13: Mark indexes as invalid when deadlock occurs. So fix based on latest comments from Masao and reverted change of v12 somewhat useless.
Review by michael-kun on 2013-03-01 04:02:53 AM: Review by Fujii Masao: crash at REINDEX CONCURRENTLY for index using expressions
Patch by michael-kun on 2013-03-01 04:03:48 AM: v14: Fix for indexes using expressions. Simply forgot to deparse correctly column names for expressions :)
Review by masaofujii on 2013-03-01 05:48:40 PM: Assertion failure occurs when rebuilding the GIN index with fastupdate = off.
Patch by michael-kun on 2013-03-04 11:58:42 AM: v15: Fix for gin/gist indexes, fix for index stats. Moved to swap with relfilenode.
Patch by michael-kun on 2013-03-04 01:37:41 PM: v16: Fix for db rebuilt after make installcheck, duplicate column names. Now stack is empty.
Review by michael-kun on 2013-03-05 01:37:15 PM: Review of v16 from Andres. Only typos and few things...
Patch by michael-kun on 2013-03-05 01:38:20 PM: v17: Corrections after review of v16. Improved error handling for toast and typo corrections
Patch by michael-kun on 2013-03-06 08:28:14 PM: v20: Adding some CHECK_FOR_INTERRUPTS for process safety
Review by masaofujii on 2013-03-09 02:21:38 AM: Encountered a segmentation fault.
Patch by michael-kun on 2013-03-11 01:12:40 AM: v22: Fixed the segenv. Due to system columns not computed correctly
Review by michael-kun on 2013-03-11 01:13:19 AM: Incorrect behavior of ALTER TABLE SET TABLESPACE
Patch by michael-kun on 2013-03-11 01:14:39 AM: v23: Fixed the pb with SET TABLESPACE. Corrected docs.
Patch by michael-kun on 2013-03-17 12:28:19 PM: v24: improvement of code and better docs
Patch by michael-kun on 2013-06-06 04:33:07 AM: v26: Realigned with master, correct comments + lock at concurrent index creation. Context diffs are used.
Review by andresfreund on 2013-06-18 10:09:12 AM: Some review comments for 20130617_1_remove_reltoastidxid_v10.patch
Review by andresfreund on 2013-06-18 12:55:08 PM: Performance testing with positive result.
Review by masaofujii on 2013-06-18 03:39:56 PM: Another review comments for 20130617_1_remove_reltoastidxid_v10.patch
Patch by michael-kun on 2013-06-19 12:57:22 AM: v27: Previous comments are all addressed
Patch by michael-kun on 2013-06-25 12:18:36 AM: v13 of reltoastidxid removal patch: looks OK for Andres. But Masao?
Review by masaofujii on 2013-06-25 04:08:26 PM: With the latest patch, VACUUM FULL failed.
Patch by michael-kun on 2013-07-01 12:54:44 AM: v15 of toast patch: Fixed VACUUM FULL, pg_upgrade (binary pg_dump)
Comment by masaofujii on 2013-07-03 06:28:31 PM: reltoastidxid removal patch has been committed.
Comment by masaofujii on 2013-07-04 05:58:44 PM: Waiting for the patch based on MVCC catalog.
Patch by michael-kun on 2013-07-05 05:04:09 AM: v28: MVCC catalogs used, swap lock lowered at ShareUpdateExclusive, wait step at swap, isolation tests and some bug fixes. 2 extra patches are provided with some infrastructure.
Comment by maiku41 on 2013-07-11 05:24:06 PM: Testing on this patch won't be complete by the end of CF due to the changes required by the just-committed MVCC catalog patch, per reviewer.
Review by andresfreund on 2013-09-16 03:42:22 PM: Swap/Drop algorithm need to be changed or it needs to be documented why they are correct despite the raised concerns.
Patch by michael-kun on 2013-11-15 03:00:27 AM: New patch rebased, labelled 2.0
Comment by petere on 2013-11-15 06:55:48 PM: needs rebase
Patch by michael-kun on 2013-11-18 12:36:58 PM: Some documentation added
Review by andresfreund on 2013-11-18 01:35:15 PM: Open concurrency issues
Comment by michael-kun on 2013-11-27 12:53:11 AM: Won't be able to allocate time fot that by the end of the CF. It would have been good to see the 1st 2 patches committed...

Add Comment

Please log in to comment on this patch.