DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=message-id:date:from:to:subject:cc:in-reply-to:mime-version
	:content-type:content-transfer-encoding:content-disposition
	:references;
	b=KV/VJLnPMswnyZkwJH3syRP4DVJS7sJJDMV1tE5X0lcZQBmYd7eAKaKLXT2fIUHzRD
	TJZfY5j7QSHCW6S3SsuFrDZzbkGjNGk7mfA5hzq8APBsA6dAyw6TM5cfaxNlg9nbjRwZ
	KGEzWAvU9MAuPP67wZOBf3OGJIWHyz4two0sE=
Message-ID: <dcc563d10809291026h6d01e20cv1afcffd8aec332c5@mail.gmail.com>
Date: Mon, 29 Sep 2008 11:26:22 -0600
From: "Scott Marlowe" <scott.marlowe@gmail.com>
To: "Steve Crawford" <scrawford@pinpointresearch.com>
Subject: Re: Do we need vacuuming when tables are regularly dropped?
Cc: "Peter Kovacs" <maxottovonstirlitz@gmail.com>,
	"Tom Lane" <tgl@sss.pgh.pa.us>, pgsql-admin@postgresql.org
In-Reply-To: <48E10C95.2050207@pinpointresearch.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <b6e8f2e80809290237p7685dbcfm9d5c886ba225b956@mail.gmail.com>
	<9815.1222690583@sss.pgh.pa.us>
	<b6e8f2e80809290533g2a7e09b1i110bb085ad75092e@mail.gmail.com>
	<48E10C95.2050207@pinpointresearch.com>

On Mon, Sep 29, 2008 at 11:12 AM, Steve Crawford
<scrawford@pinpointresearch.com> wrote:
>
>>> What it sounds like to me is that you're not vacuuming the system
>>> catalogs, which are getting bloated with dead rows about all those
>>> dropped tables.
>>>
>>
>> Wow, great!
>>
>> It is not immediately clear from the documentation, but the VACUUM
>> command also deals with the system catalogs as well, correct?
>>
>>
>
> To expand on Tom's answer, rows in system tables are created not only for
> tables but for each column in the table, rules, indexes, etc. You  can end
> up with a lot more row creation than you suspect. And temporary tables bloat
> the system tables just like regular tables. We discovered that cron scripts
> using temporary tables can cause very rapid system-table blotage.

Also, there was a time when you couldn't do vacuum full on system
tables do to locking issues, and had to take the db down to single
user mode to do so.

Tom, is that still the case?