fwiw, we have a lot of materialized views, so restoring a DB on non-vacuumed tables caused the materialization to take a lot longer than it would have with autovacuum running as normal. Seems worth experimenting though.
> On Jul 28, 2024, at 6:40 AM, Laurenz Albe <laurenz.albe@cybertec.at> wrote: > > That's bad advice. Very bad advice. > That is, unless you are ready to delete the cluster and run a new "initdb" after an OS crash.
Exactly.
> You are wrong: it is not the database that is broken after a crash, but the entire cluster.
Good clarification. I personally have never had occasion to move a partial cluster, so my use of "database" in my question was sloppy, I meant "cluster". So yes, I'd delete the cluster and initdb if I ever actually had an OS crash during a pg_restore--which in 20 years of using PG has never happened. I suppose it might matter more if one were forced to run one's db on an unstable platform ;-)