On Mar 14, 2006, at 1:31 PM, Casey Duncan wrote:
> On Mar 13, 2006, at 5:25 PM, Casey Duncan wrote:
> [..]
>>>> If I restart the postmaster, the query will complete in the expected
>>>> time.
>>>
>>> Does the problem eventually start happening again? If so, after
>>> how long? How did you determine that the restart is relevant? Do
>>> you consistently see different (presumably better) performance after
>>> a restart than if you don't restart?
>
> I restarted postgres this morning and this time it didn't seem to
> help. That query has been running for several hours now. I'm going to
> let it go a while longer to see if it eventually completes, but I
> suspect it won't. So perhaps the prior restart was just dumb luck.
>
> I'll try some narrower queries as well.
Interestingly, I can only narrow the query by about an hour (i.e., a 23
hour span) before the plan changes to a simple index scan. That may
explain why this only started hanging recently since it probably only
started using the bitmap scan for a day interval recently.
Let me know if there is anything I should do or any info you need to
further pin this down.
In the mean time I will disable bitmap scans for this query.
Thanks.
-Casey