Matthew Schumacher <matt(dot)s(at)aptalaska(dot)net> writes:
> I have been having performance problems with my DB so this morning I
> added some config to log queries that take more than 250ms. The result
> is surprising because some queries will take as long as 10 seconds, but
> then you do a explain analyze on them they show that indexes are being
> used and they run very fast.
Is it possible that it's not directly that query's fault? For instance
it could be blocked by a lock held by some other transaction. I can't
unfortunately think of any very nice way to deduce this from log entries
... you'd have to catch it in the act and look into pg_locks to find out
who's the perpetrator.
regards, tom lane