| From: | Michael Lewis <mlewis(at)entrata(dot)com> |
|---|---|
| To: | Thomas Kellerer <shammat(at)gmx(dot)net> |
| Cc: | PostgreSQL General <pgsql-general(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: Accounting for between table correlation |
| Date: | 2021-01-15 18:28:39 |
| Message-ID: | CAHOFxGq4zH4CYX4BOMH3O5a0=J3oWyJbYJU6USxwOgHubok-VA@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-general |
On Fri, Jan 15, 2021 at 9:10 AM Thomas Kellerer <shammat(at)gmx(dot)net> wrote:
> Maybe there are many "idle in transaction" sessions that prevent
> autovacuum from properly
> analyzing those tables. Or maybe for some unknown reason autovacuum was
> turned off.
> Or maybe they receive a lot of bulk loads which would require a manual
> analyze.
>
"hundreds of millions of rows"
Also of note that the default autovacuum settings for scale factor of
10-20% for vacuum & analyze behavior may need to be lowered for such a
table. OP has chimed in that they believe it is being vacuumed/analyzed
often enough, but if data is primarily or onnly maintained by bulk load, it
would seem like autovacuum/analyze wouldn't be a factor anyway.
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Michael Lewis | 2021-01-15 18:49:04 | Re: time-based range partitioning and truncate/delete different timezones |
| Previous Message | Michael Lewis | 2021-01-15 18:27:26 | Re: Accounting for between table correlation |