On Wed, 07 Mar 2007 09:27:04 -0800
Reece Hart <reece@harts.net> wrote:
> On Wed, 2007-03-07 at 10:08 -0600, John Gateley wrote:
> > I'm guessing it's something related to table locks.
> ...
> > Any pointers on what I should be looking for to prevent this from
> > happening again? What information I should be tracking to figure
> > out what is exactly happening?
>
> Your inserts almost certainly have a table or index exclusively locked
> and thereby causing a backlog of selects.
Thanks. It turns out it was my nightly vacuuming of the database.
I had the full option set, and I had added a large table, and I
think it was just trying to finish the vacuum (it took about
15 minutes to do a vacuumdb -az, would several hours be reasonable
for afz? Or maybe there was some deadlock with table access?)
Moral - full isn't always better (and if I RTFM, I would have
known that...)
j