On 7/31/07, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Merlin Moncure" <mmoncure@gmail.com> writes:
> > third (but unlikely) possibility is there are various dropped tables,
> > etc which need to be deleted but there are stale postgresql processes
> > holding on to the fd. This would only happen following a postmaster
> > crash or some other bizarre scenario, but i've seen it on production
> > box.
>
> Brent Reid reported something similar in bug #3483 but I'm still quite
> unclear how it'd happen in any realistic scenario. Can you create a
> test case?
No, but I've seen it on a production 8.1 box (once). I didn't
actually cause the problem, just cleaned it up. It was unnoticed for
several weeks/months because the postmaster processes showed up
without a controlling tty.
My best guess is the postmaster was killed improperly out of haste
during a maintenance window, or possibly an out of disk space related
issue at an earlier point. I never really considered that it was a
postgresql problem.
merlin