From: | Mischa Sandberg <mischa(at)ca(dot)sophos(dot)com> |
---|---|
To: | Andrew Sullivan <ajs(at)crankycanuck(dot)ca> |
Cc: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: vacuuming problems continued |
Date: | 2006-06-06 18:51:56 |
Message-ID: | 4485CECC.4040207@ca.sophos.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
Andrew Sullivan wrote:
> On Thu, Jun 01, 2006 at 01:54:08PM +0200, Antoine wrote:
>> Hi,
>> We just don't seem to be getting much benefit from autovacuum. Running
>> a manual vacuum seems to still be doing a LOT, which suggests to me
>> that I should either run a cron job and disable autovacuum, or just
>> run a cron job on top of autovacuum.
Don't know if this was covered in an earlier thread. Bear with me if so.
I'm working with 7.4.8 and 8.0.3 systems, and pg_autovacuum does have some
glitches ... in part solved by the integrated autovac in 8.1:
- in our env, clients occasionally hit max_connections. This is a known and
(sort of) desired pushback on load. However, that sometimes knocks pg_autovacuum
out.
- db server goes down for any reason: same problem.
Just restarting pg_autovacuum is not good enough; when pg_autovacuum terminates,
it loses its state, so big tables that change less than 50% between such
terminations may never get vacuumed (!)
For that reason, it's taken a switch to a Perl script run from cron every 5
minutes, that persists state in a table. The script is not a plug-compatible
match for pg_autovacuum (hardcoded rates; hardcoded distinction between user and
system tables), but you may find it useful.
--
Engineers think that equations approximate reality.
Physicists think that reality approximates the equations.
Mathematicians never make the connection.
Attachment | Content-Type | Size |
---|---|---|
autovac | text/plain | 4.8 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Joshua D. Drake | 2006-06-06 19:21:39 | Re: vacuuming problems continued |
Previous Message | Scott Marlowe | 2006-06-06 17:54:27 | Re: Some queries starting to hang |