| From: | Jack Christensen <jack(at)jackchristensen(dot)com> |
|---|---|
| To: | pgsql <pgsql-general(at)postgresql(dot)org> |
| Subject: | Is auto-analyze as thorough as manual analyze? |
| Date: | 2017-09-23 22:10:42 |
| Message-ID: | 5ddd2617-66b0-a6e2-3af4-618e09a96647@jackchristensen.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-general |
Just had an issue where a prepared query would occasionally choose a
very bad plan in production. The same data set in a different
environment consistently would choose the index scan. As would be
expected, running analyze on that table in production resolved the issue.
However, before I ran the analyze I checked pg_stat_user_tables to see
last_autoanalyze for that table. It had run today. But the problem
existed before that. I would have expected that the auto-analyze would
have corrected this (or prevented it entirely if run enough).
So that leaves me wondering: is an auto-analyze the same as manually
running analyze or is a manual analyze more thorough? This is running
version 9.6.3 on Heroku.
Thanks,
Jack
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Melvin Davidson | 2017-09-23 22:48:17 | Re: Is auto-analyze as thorough as manual analyze? |
| Previous Message | Paul A Jungwirth | 2017-09-23 17:09:08 | Re: Is float8 a reference type? |