Re: Autovacuum daemon internal handling

From: Alvaro Herrera <alvherre(at)commandprompt(dot)com>
To: Gnanakumar <gnanam(at)zoniac(dot)com>
Cc: pgsql-admin <pgsql-admin(at)postgresql(dot)org>
Subject: Re: Autovacuum daemon internal handling
Date: 2010-08-11 15:17:28
Message-ID: 1281539491-sup-7017@alvh.no-ip.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Excerpts from Gnanakumar's message of mié ago 11 01:07:08 -0400 2010:

Hi,

> 1. Does autovacuum daemon works with one table at a time or does it work
> with multiple tables at the same time?

Only one.

> Reason to ask this question is, let's say I've "autovacuum_naptime" set to 1
> minute, and there are 500 tables in one database. Assuming there are 10
> tables that has to be vacuumed and analyzed based on the threshold defined
> in one of the autovacuum daemon execution time, and if takes more than 1
> minute to complete vacuum and analyze operation on even a single table (out
> of 10 tables), how does autovacuum daemon works/handles in this case?

in 8.2 "naptime" means "time to sleep after we finish a job". So even
if the previous task takes an hour, it will still sleep a minute before
doing another round. (Note that this setting has a different meaning in
later releases).

If autovacuum cannot keep up with all the vacuumable tables, you're in
trouble and should probably schedule vacuum externally. (This also
changed in later releases).

> 2. The columns "last_autovaccum" and 'last_autoanalyze" in
> pg_stat_user_tables shows the start time or end time of the operation?

End time.

--
Álvaro Herrera <alvherre(at)commandprompt(dot)com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

In response to

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Kevin Grittner 2010-08-11 15:27:11 Re: Autovacuum daemon internal handling
Previous Message Robert Haas 2010-08-11 14:25:58 Re: [HACKERS] postgres 9.0 crash when bringing up hot standby