Hi, moving this to -hackers.
On Wed, May 27, 2026 at 12:06:45PM +0300, Heikki Linnakangas wrote: > On 27/05/2026 05:55, Nazneen Jafri wrote: > > Tested Andrey's demo.diff on a fresh environment: > > > > - Primary: REL_16_8, Standby: REL_16_14 (--enable-cassert) > > > > - ~2300 MultiXacts crossing the offsets page boundary > > > > - Without patch: startup deadlocks at RecordNewMultiXact(multi=2047) > > > > - With patch: standby replays all WAL and catches up > > Thanks all. I have applied this to v14 - v16. I would like to suggest that this bug-fix might warrant an out-of-cylce release (even though it is has been a month since the fix got committed). There has been at least one more report on pgsql-bugs[1], we (at credativ) have had a customer issue about it and Jan Karremans (in CC) mentioned this on Telegram today as well. To reiterate, this bug is a regression in the latest back-branch releases for versions 14-16. If either the standby gets updated to the latest minor release first, or if somebody does PITR from a primary on an earlier release to an instance with the latest minor release, the standby/PITR deadlocks and just sits there. I think those are pretty important/usual use-cases, so even if versions 17/18 are not affected (to-the-best-of-my-knowledge), maybe 14-16 should get a re-release for this? Or has this already been discussed among the release team? Regards, Michael
