I think I can answer this cause I recently had a similar problem. There is a
voodoo setting in postgresql called "analyze target". It controls how much
statistic information is kept per table. This information affects the query
planner. If it makes a bad guess based on insufficient statistics data, it will
absolutely kill performance (BTW, the documentation never makes it explicit).
Increase default_analyze_target (I think that's what it's called, look up the
docs) at least tenfold, restart postgresql, and run analyze again.
BTW, the default postgresql settings are WAY too conservative. I am now looking
into tuning and there are a lot of things that need to be turned up.
hope this helps,
Eugene
>
> From: Michael Fuhr <[EMAIL PROTECTED]>
> Date: 2005/08/18 Thu AM 10:05:14 EST
> To: WireSpot <[EMAIL PROTECTED]>
> CC: pgsql-general@postgresql.org
> Subject: Re: [GENERAL] Same database, different query plans
>
> On Thu, Aug 18, 2005 at 12:03:59PM +0300, WireSpot wrote:
> > The actual SELECT results (ie. non EXPLAIN) are identical in both
> > cases. The indexes and so on are identical. I've done a reindexing and
> > vacuuming on both of them just to be sure.
> >
> > As you can see, there's quite a bit of a difference between 0.3 ms and
> > 398 ms, and it shows. I haven't touched the query planning options.
> > Why the different planning and what can I do to fix the misguided one?
>
> Have you run ANALYZE or VACUUM ANALYZE in both databases to update
> the planner's statistics? If you have and get the same results,
> then it might be interesting to see the output of the following on
> both systems:
>
> SET enable_mergejoin TO off;
> SET enable_nestloop TO on;
> EXPLAIN ANALYZE SELECT ...
>
> SET enable_mergejoin TO on;
> SET enable_nestloop TO off;
> EXPLAIN ANALYZE SELECT ...
>
> --
> Michael Fuhr
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: explain analyze is your friend
>
---------------------------(end of broadcast)---------------------------
TIP 1: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to [EMAIL PROTECTED] so that your
message can get through to the mailing list cleanly
