Philip Warner <pjw@rhyme.com.au> writes:
> At 02:33 PM 3/05/2004, Tom Lane wrote:
>> Could you dig a little deeper and see where the problem really is?
> I thought I had 8-(.
> The result of a 'select * from pg_locks where not granted' was a bunch of
> locks on the pg_listener relation, and no others. Only one process had a
> lock on that relation, and it was an ANALYZE command.
I don't believe any of this. In the first place, if you're not using
LISTEN/NOTIFY then there's no reason for any backend to try to take out
an AccessExclusive lock on pg_listener (which is the only kind that would
be blocked by ANALYZE's measly AccessShareLock). In the second place,
an ANALYZE on a zero-page relation cannot conceivably take half an hour,
unless it's in turn being blocked by something else.
Please dig deeper.
regards, tom lane