Thread: Re: [PATCHES] tiny patch to make vacuumdb -a's database order match pg_dumpall

"Dan Thomas" <godders@gmail.com> writes:
> I've been having trouble running vacuumdb -a and pg_dumpall
> concurrently because they run through the databases in a different
> order (so dumpall was getting stuck behind vacuum's lock, and my
> firewall was rather unhelpfully closing the idle connection).

Um, whaddya mean "dumpall was getting stuck behind vacuum's lock"?
A plain vacuum doesn't take any locks that would block pg_dump.

While the proposed patch looks harmless enough, I'm unconvinced that
it will solve your problem, or even quite what the problem is.

            regards, tom lane

> Um, whaddya mean "dumpall was getting stuck behind vacuum's lock"?
> A plain vacuum doesn't take any locks that would block pg_dump.

Dammit, just looked and the setup I originally encountered the problem
on and tracked it down to the vacuum process, and it is indeed set up
to perform a full vacuum.. I've incorrectly assumed the problem I'm
having now (with a normal vac) was for the same reason.

> While the proposed patch looks harmless enough, I'm unconvinced that
> it will solve your problem, or even quite what the problem is.

Yes, sorry about that, it does indeed appear that whatever is causing
my dumpall process to die isn't PG's fault.

Though I still think it makes a *bit* of sense to have vacuumdb use
the same order as pg_dumpall (clusterdb too now I think about it),
it's obviously not as much of an issue as I originally thought, and
not the source of my problem, which is a shame :)

Dan