To: Ken Caruso <ken@ipl31.net>
cc: pgsql-admin@postgresql.org
Subject: Re: Bloat and Slow Vacuum Time on Toast 
In-reply-to: 
 <CAMg8r_o=sHXqOzPgNw9+ndp+-0bAK6DBcjCu0poZn2se5qC-vQ@mail.gmail.com>
References: 
 <CAMg8r_o=sHXqOzPgNw9+ndp+-0bAK6DBcjCu0poZn2se5qC-vQ@mail.gmail.com>
Comments: In-reply-to Ken Caruso <ken@ipl31.net>
	message dated "Tue, 19 Jul 2011 13:37:47 -0700"
Date: Tue, 19 Jul 2011 17:26:50 -0400
Message-ID: <5036.1311110810@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Ken Caruso <ken@ipl31.net> writes:
> I am trying to trouble shoot a couple of issue related to vacuum and db
> size. I currently have a postgres DB that is around 390GB on disk (du -hs in
> $PG_DATA/base). If I look at the total relation size using the following
> query:

> SELECT SUM(pg_relation_size(pg_class.oid))
> FROM pg_class ;

> This says the total size is around 191GB. That query should sum all of the
> tables, toast and indexes AFAIK.

What PG version?

If it's a version with multiple table forks, you're only counting the
main forks, not FSM or VM.

Theoretically the latter two should be pretty small relative to the main
fork, but obviously you are missing something with this query ...

			regards, tom lane