Обсуждение: Why auto vacuum almost running all the time on one toast table?

v9.1.3 on Ubuntu 10.04 server. I have one table which has frequent
insert and weekly deletion, no update. Recently, I found the auto vacuum
on the toast table of it almost running all the time. each run took
around 1 hour. I don't think there are so many inerts so that it reaches
the auto vacuum criteria(20%? changes). And there is no such frequent
auto vacuum on the main table. The last auto vacuum of the main table is
on 3rd Jun. what might cause this behavior?

# grep pg_toast_16922 postgresql-2012-06-12_000000.log
2012-06-12 01:22:56 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"
2012-06-12 04:27:49 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"
2012-06-12 06:58:02 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"
2012-06-12 09:27:38 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"
2012-06-12 12:17:33 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"
2012-06-12 15:05:31 CST [@] CONTEXT:  automatic vacuum of table
"oodb.pg_toast.pg_toast_16922"

Rural Hunter <ruralhunter@gmail.com> writes:
> v9.1.3 on Ubuntu 10.04 server. I have one table which has frequent
> insert and weekly deletion, no update. Recently, I found the auto vacuum
> on the toast table of it almost running all the time. each run took
> around 1 hour. I don't think there are so many inerts so that it reaches
> the auto vacuum criteria(20%? changes). And there is no such frequent
> auto vacuum on the main table. The last auto vacuum of the main table is
> on 3rd Jun. what might cause this behavior?

Hm, is the autovacuum managing to complete, or is it getting kicked off
before it can complete?  It would help to see the messages those
CONTEXT lines are attached to; and/or you might check pg_stat_all_tables
to see when the last completed autovacuum was for both the main table
and its toast table.

            regards, tom lane