Re: A failure in 031_recovery_conflict.pl on Debian/s390x

Christoph Berg <myon@debian.org>

From: Christoph Berg <myon@debian.org>

To: Andres Freund <andres@anarazel.de>

Cc: Thomas Munro <thomas.munro@gmail.com>, Noah Misch <noah@leadboat.com>, Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Melanie Plageman <melanieplageman@gmail.com>

Date: 2023-08-28T13:58:01Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Fix instability in 031_recovery_conflict.pl.
- 0174c2d213f6 17.0 landed
Fix recovery conflict SIGUSR1 handling.
- 0da096d78e1e 17.0 landed
Disable 031_recovery_conflict.pl in 15 and 16.
- e13de4913913 15.5 landed
- 8d1cf9674a25 16.0 landed

Re: Andres Freund
> > Thanks.  I realised that it's easy enough to test that theory about
> > cleanup locks by hacking ConditionalLockBufferForCleanup() to return
> > false randomly.  Then the test occasionally fails as described.  Seems
> > like we'll need to fix that test, but it's not evidence of a server
> > bug, and my signal handler refactoring patch is in the clear.  Thanks
> > for testing it!
> 
> WRT fixing the test: I think just using VACUUM FREEZE ought to do the job?
> After changing all the VACUUMs to VACUUM FREEZEs, 031_recovery_conflict.pl
> passes even after I make ConditionalLockBufferForCleanup() fail 100%.

I have now applied the last two patches to postgresql-17 so see if the
build is more stable. (So far I had only tried in manual tests.)

Fwiw this is also causing pain on PostgreSQL 16:

https://pgdgbuild.dus.dg-i.net/view/Snapshot/job/postgresql-16-binaries-snapshot/1011/architecture=s390x,distribution=sid/consoleText

Most of the failing builds in
https://pgdgbuild.dus.dg-i.net/view/Snapshot/job/postgresql-16-binaries-snapshot/
are on s390x and likely due to this problem.

This should be fixed before the 16 release.

Christoph