Re: Re: Vaccuuming every db in cluster fails to eliminate "execute a full-database VACUUM in..."

Robert Burgholzer <rburghol(at)vt(dot)edu> writes:
> So, as I mentioned, I have tried to get this straightened out by
> vacuuming all in --single mode, but to no avail. I executed the
> following command, to see which tables were in trouble:

> SELECT relname, age(relfrozenxid) FROM pg_class WHERE relkind = 'r';

> And found a table listed, that DOES NOT exist any longer.

Yeah? What happens if you try to select from that table?

If there is a row in pg_class that for some reason didn't get deleted
when the table was dropped, you could just manually remove that row
(ie, DELETE FROM pg_class WHERE ... as superuser). You'd still need
another VACUUM to get the database's datfrozenxid updated, but
after that things should be OK.

regards, tom lane