From: "Anibal David Acosta" <aa@devshock.com>
To: "'Kevin Grittner'" <Kevin.Grittner@wicourts.gov>,
	<pgsql-performance@postgresql.org>
References: <009d01cd7bee$785b1380$69113a80$@devshock.com>
	<502D173702000025000497D1@gw.wicourts.gov>
In-Reply-To: <502D173702000025000497D1@gw.wicourts.gov>
Subject: Re: best practice to avoid table bloat?
Date: Thu, 16 Aug 2012 17:10:31 -0400
Message-ID: <00c201cd7bf3$93bc4130$bb34c390$@devshock.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
Thread-Index: AQK7QPGWhO6yMB3u2/1Hg2+iMZ7ltwHSQBc1lXLPf8A=
Content-Language: es-py

Thanks Kevin.
Postgres version is 9.1.4 (lastest)

Every day the table has about 7 millions of new rows.
The table hold the data for 60 days, so approx. the total rows must be
around 420 millions.
Every night a delete process run, and remove rows older than 60 days.

So, the space used by postgres should not be increase drastically because
every day arrive 7 millions of rows but also same quantity is deleted but my
disk get out of space every 4 months.
I must copy tables outside the server, delete local table and create it
again, after this process I got again space for about 4 months.

Maybe is a wrong autovacuum config, but is really complicate to understand
what values are correct to avoid performance penalty but to keep table in
good fit.

I think that autovacuum configuration should have some like "auto-config"
that recalculate every day which is the best configuration for the server
condition

Thanks!


-----Mensaje original-----
De: Kevin Grittner [mailto:Kevin.Grittner@wicourts.gov] 
Enviado el: jueves, 16 de agosto de 2012 04:52 p.m.
Para: Anibal David Acosta; pgsql-performance@postgresql.org
Asunto: Re: [PERFORM] best practice to avoid table bloat?

"Anibal David Acosta" <aa@devshock.com> wrote:
 
> if I have a table that daily at night is deleted about 8 millions of 
> rows (table maybe has 9 millions) is recommended to do a vacuum 
> analyze after delete completes or can I leave this job to autovacuum?
 
Deleting a high percentage of the rows should cause autovacuum to deal with
the table the next time it wakes up, so an explicit VACUUM ANALYZE shouldn't
be needed.
 
> For some reason for the same amount of data every day postgres consume 
> a little more.
 
How are you measuring the data and how are you measuring the space? 
And what version of PostgreSQL is this?
 
-Kevin