| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Anthony Presley <anthony(at)resolution(dot)com> |
| Cc: | pgsql-performance <pgsql-performance(at)postgresql(dot)org>, Erik Van Gilder <evg(at)resolution(dot)com> |
| Subject: | Re: 8.4.4, 9.0, and 9.1 Planner Differences |
| Date: | 2011-10-23 02:31:59 |
| Message-ID: | 20459.1319337119@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
Anthony Presley <anthony(at)resolution(dot)com> writes:
> I have tried setting the statistics on employee.user_id to be 100 and 1000,
> and the rest are the default (100).
> I've run both an "ANALYZE" and a "VACUUM ANALYZE" on the production system -
> both "generally", and on each of the above tables (employee, app_user,
> location, preference).
> Here's an updated explain of the most recent attempt. About 5 minutes after
> I analyzed them:
> http://explain.depesz.com/s/G32
Looks like the biggest estimation errors are on the location_id joins.
Maybe you should be cranking up the stats targets on those columns.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Robins Tharakan | 2011-10-23 10:38:19 | Re: explain workload |
| Previous Message | Stefan Keller | 2011-10-22 23:33:53 | hstore query: Any better idea than adding more memory? |