To: Joe Miller <joe.d.miller@gmail.com>
cc: pgsql-performance@postgresql.org
Subject: Re: Auto ANALYZE criteria 
In-reply-to: <AANLkTi=W02Lcn63qqxoP9jjQ0CJ8wZfCJkKV7fry7CP=@mail.gmail.com> 
References: <AANLkTi=W02Lcn63qqxoP9jjQ0CJ8wZfCJkKV7fry7CP=@mail.gmail.com>
Comments: In-reply-to Joe Miller <joe.d.miller@gmail.com>
	message dated "Mon, 20 Sep 2010 16:38:37 -0400"
Date: Mon, 20 Sep 2010 22:12:31 -0400
Message-ID: <21216.1285035151@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Joe Miller <joe.d.miller@gmail.com> writes:
> The autovacuum daemon currently uses the number of inserted and
> updated tuples to determine if it should run VACUUM ANALYZE on a
> table.  Why doesn’t it consider deleted tuples as well?

I think you misread the code.

Now there *is* a problem, pre-9.0, if your update pattern is such that
most or all updates are HOT updates.  To quote from the 9.0 alpha
release notes:

         Revise pgstat's tracking of tuple changes to
         improve the reliability of decisions about when to
         auto-analyze.  The previous code depended on n_live_tuples +
         n_dead_tuples - last_anl_tuples, where all three of these
         numbers could be bad estimates from ANALYZE itself.  Even
         worse, in the presence of a steady flow of HOT updates and
         matching HOT-tuple reclamations, auto-analyze might never
         trigger at all, even if all three numbers are exactly right,
         because n_dead_tuples could hold steady.

It's not clear to me if that matches your problem, though.

			regards, tom lane