Re: Missing the point of autovacuum

From: Tobias Brox <tobias(at)nordicbet(dot)com>
To: Daryl Herzmann <akrherz(at)iastate(dot)edu>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: Missing the point of autovacuum
Date: 2007-01-06 19:33:13
Message-ID: 20070106193313.GB30486@oppetid.no
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

[Daryl Herzmann - Sat at 12:59:03PM -0600]
> As the months have gone by, I notice many of my tables having *lots* of
> unused item pointers. For example,

Probably not the issue here, but we had some similar issue where we had
many long-running transactions - i.e. some careless colleague entering
"begin" into his psql shell and leaving it running for some days without
entering "commit" or "rollback", plus some instances where the
applications started a transaction without closing it.

> Perhaps I shouldn't be concerned with this? In all, I've got around 400
> GB of data on postgresql, but am not sure how much of it is old data.

I didn't count the zeroes, but autovacuum does have rules saying it will
not touch the table until some percentages of it needs to be vacuumed
off. This is of course configurable.

> Do I need to be running old fashioned 'vacuumdb' routinely as well? I
> guess I just don't understand why autovacuum is not automatically doing
> this for me and I have tables with so many unused item pointers.

If you have some period of the day with less activity than else, it is a
good idea running an old-fashionated vacuum as well. The regular vacuum
process will benefit from any work done by the autovacuum.

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Tom Lane 2007-01-06 19:46:40 Re: Missing the point of autovacuum
Previous Message Daryl Herzmann 2007-01-06 18:59:03 Missing the point of autovacuum