On Thu, 2023-05-18 at 17:34 +0100, Owen Stephens wrote:
> We are seeing that vacuum is prevented from cleaning dead tuples by an open
> transaction in a different database (where both connections are made against the
> primary server) when hot_standby_feedback = on but not when it is off. Is this
> cross-database interaction an expected effect of enabling hot_standby_feedback,
> even if the connections interact only with the primary not the replica?
Yes, that's what I would expect. There is only one "backend_xmin" in
"pg_stat_replication", which corresponds to the snapshot held by the oldest
query in any database on the standby server.
Yours,
Laurenz Albe