On Tue, Aug 2, 2022 at 12:32 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Hmmm ... now that you mention it, I see nothing in 002_pg_upgrade.pl
> that attempts to turn off autovacuum on either the source server or
> the destination. So one plausible theory is that autovac moved the
> numbers since we checked.
It's very easy to believe that my work in commit 0b018fab could make
that happen, which is only a few months old. It's now completely
routine for non-aggressive autovacuums to advance relfrozenxid by at
least a small amount.
For example, when autovacuum runs against either the tellers table or
the branches table during a pgbench run, it will now advance
relfrozenxid, every single time. And to a very recent value. This will
happen in spite of the fact that no freezing ever takes place -- it's
just a consequence of the oldest extant XID consistently being quite
young each time, due to workload characteristics.
--
Peter Geoghegan
