Horton

Re: BUG #17055: Logical replication worker crashes when applying update of row that dose not exist in target partiti

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Amit Langote <amitlangote09@gmail.com>, sbernikov@gmail.com, PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Date: 2021-06-12T01:49:04Z
Lists: pgsql-bugs 
On Fri, Jun 11, 2021 at 09:37:05PM -0400, Tom Lane wrote:
> If you prefer.  However, I'd note that this patch as it stands
> sort of breaks the argument I just made, as now you are interrogating
> the postmaster's state after sending SIGHUP and before doing anything
> that will wait for that to take effect.  I think you should move the
> location-capture commands down by at least one safe_psql call.

Ah, right.  That would be after "DELETE FROM tab_full_pk" in
001_rep_changes.pl and after "DELETE FROM tab1" in 001_rep_changes.pl
run on the subscribers, then.  Would you prefer doing the change?  If
you don't think that's not worth it, I won't fight hard on it, but
while we're dealing with it..
--
Michael

Commits

  1. Improve log pattern detection in recently-added TAP tests

  2. Fix multiple crasher bugs in partitioned-table replication logic.