Horton

Re: failures in t/031_recovery_conflict.pl on CI

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-hackers@postgresql.org, Thomas Munro <thomas.munro@gmail.com>
Date: 2022-04-12T18:49:13Z
Lists: pgsql-hackers 
Hi,

On 2022-04-09 19:34:26 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > It's been broken in different ways all the way back to 9.0, from what I can
> > see, but I didn't check every single version.
> 
> > Afaics the fix is to nuke the idea of doing anything substantial in the signal
> > handler from orbit, and instead just set a flag in the handler.
> 
> +1.  This is probably more feasible given the latch infrastructure
> than it was when that code was first written.

What do you think about just reordering the disable_all_timeouts() to be
before the got_standby_deadlock_timeout check in the back branches? I think
that should close at least the most obvious hole.  And fix it properly in
HEAD?

Greetings,

Andres Freund

Commits

  1. Disable 031_recovery_conflict.pl until after minor releases.

  2. Temporarily skip recovery deadlock test in back branches.

  3. Backpatch addition of pump_until() more completely.

  4. Fix timing issue in deadlock recovery conflict test.

  5. Backpatch 031_recovery_conflict.pl.

  6. Fix possibility of self-deadlock in ResolveRecoveryConflictWithBufferPin().

  7. Backpatch addition of wait_for_log(), pump_until().

  8. Add tests for recovery deadlock conflicts.

  9. Phase 3 of pgindent updates.