At Wed, 28 Feb 2024 03:23:37 +0000, Mark Schloss <Mark.Schloss@austrac.gov.au> wrote in
> Thank you for your reply. I can confirm there were no changes made to the config of the replica.
Okay.
> Is there any significance in the parameters in the commit record -
> 'inval msgs: catcache 21; sync'.
I think it is not relevant.
> - The walreceiver on the barman server did not fail but the WAL file does not contain the commit transaction
I don't have detailed knowledge of barman's behavior, but it seems to
be quite normal for barman to write out only on receiving a commit
record. What I don't understand here is how those WAL files on the
barman server are related to the failed replicas.
From the information you provided, I guess that the replicas somehow
obtain the currently-written WAL file from the barman server at a
certain time through a route unknown to me, but you seem to know
that. I think that relationship has not been explained here.
Could you explain the routes and timings that WAL files are copied
between the servers?
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center