To: David Ondrejik <David.Ondrejik@noaa.gov>
cc: Simon Riggs <simon@2ndQuadrant.com>,
	pgsql-admin <pgsql-admin@postgresql.org>
Subject: Re: vacuumdb question/problem 
In-reply-to: <4E208B57.6010401@noaa.gov> 
References: <4E206658.4040600@noaa.gov>
	<CA+U5nMLRcZRceB9FMeJBJN3fPyqNeGmEapzSXS3h=omPyntJeg@mail.gmail.com>
	<3239.1310749967@sss.pgh.pa.us> <4E208B57.6010401@noaa.gov>
Comments: In-reply-to David Ondrejik <David.Ondrejik@noaa.gov>
	message dated "Fri, 15 Jul 2011 14:47:51 -0400"
Date: Fri, 15 Jul 2011 15:11:11 -0400
Message-ID: <17933.1310757071@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

David Ondrejik <David.Ondrejik@noaa.gov> writes:
> I am still wondering how the vacuum process actually works. When it 
> throws the output lines that show how many rows are 
> recoverable/nonremovable, does this mean that the vacuum has completed?  

No, that's just the end of the first pass over the table.  After that it
cleans out dead index entries, and then it sets about moving rows down
to fill free space.

As far as your original question is concerned: if you have reason to
think there's a lot of bloat, CLUSTER would be the thing to use.
Old-style VACUUM FULL is seriously inefficient when there's a lot of
rows that have to be moved.

			regards, tom lane