From: Chris Browne <cbbrowne@acm.org>
Subject: Re: 8.2 Autovacuum BUG ?
Date: Fri, 31 Aug 2007 15:13:54 -0400
Organization: cbbrowne Computing Inc
Lines: 43
Message-ID: <60wsvbcya5.fsf@dba2.int.libertyrms.com>
References: <46D83427.60001@livedatagroup.com>
 <25038.1188574775@sss.pgh.pa.us>
	<46D83C50.3040409@livedatagroup.com>
	<20070831161158.GD8046@alvh.no-ip.org>
	<46D840EC.1060302@livedatagroup.com>
	<1188577771.22730.238.camel@archimedes>
	<46D86103.6060204@livedatagroup.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.1008 (Gnus v5.10.8) XEmacs/21.4.20 (linux)
Cancel-Lock: sha1:46JrHeLFCNcdvwla9CI+UpQpdgk=
To: pgsql-performance@postgresql.org

pkalva@livedatagroup.com (Pallav Kalva) writes:
> Mark Lewis wrote:
>> On Fri, 2007-08-31 at 12:25 -0400, Pallav Kalva wrote:
>>
>>> Can you please correct me if I am wrong, I want to understand how
>>> this works.
>>> Based on what you said, it will run autovacuum again when it passes
>>> 200M transactions, as SELECTS are transactions too and are going on
>>> these tables.
>>> But the next time when it runs autovacuum, it shouldnt freeze the
>>> tuples again as they are already frozen and wont generate lot of
>>> archive logs ?
>>> Or is this because of it ran autovacuum for the first time on this
>>> db ? just the first time it does this process ?
>>>
>>
>> That is correct.  The tuples are now frozen, which means that they will
>> not need to be frozen ever again unless you insert/update any records.
>>
>>
>
> My main concern is filling up my disk with archive logs, so from all
> the replies I get is that since tuples are already frozen, next time
> when it runs autovacuum it wont generate any archive logs.
>
> Is my assumption right ?

No, your assumption is wrong.

Later vacuums will not generate archive files for the tuples that were
*previously* frozen, but if you have additional tuples that have
gotten old enough to reach the "freeze point," THOSE tuples will get
frozen, and so you'll continue to see archive logs generated.

And this is Certainly Not A Bug.  If the system did not do this, those
unfrozen tuples would eventually disappear when your current
transaction XID rolls over.  The freezing is *necessary.*
-- 
let name="cbbrowne" and tld="cbbrowne.com" in name ^ "@" ^ tld;;
http://linuxdatabases.info/info/unix.html
Rules of  the Evil Overlord  #86. "I will  make sure that  my doomsday
device is up to code and properly grounded."
<http://www.eviloverlord.com/>