Re: Subscription test 013_partition.pl fails under CLOBBER_CACHE_ALWAYS

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: pgsql-hackers@lists.postgresql.org
Date: 2020-09-16T00:16:26Z
Lists: pgsql-hackers
I wrote:
> With this, we get through 013_partition.pl under CCA.  I plan to
> try to run all of subscription/ and recovery/ before concluding
> there's nothing else to fix, though.

Looks like the rest passes.  FTR, it was possible to get through
subscription/ in about 2 hours on my workstation, and recovery/ in about
30 minutes, after hacking things so that initdb used a non-CCA backend.
So that's definitely a promising thing to look at if anyone wants
to try to do this on a regular basis.

Aside from increasing wal_receiver_timeout as previously mentioned,
I found I had to set a higher pg_ctl start timeout using PGCTLTIMEOUT
in order to get through the recovery tests.

			regards, tom lane



Commits

  1. Fix bogus cache-invalidation logic in logical replication worker.