From: Marc Cousin <mcousin@sigma.fr>
Organization: Sigma Informatique
To: "Francisco Reyes" <lists@stringsutils.com>
Subject: Re: Inserts optimization?
Date: Fri, 14 Apr 2006 14:00:38 +0200
User-Agent: KMail/1.9.1
Cc: "Chris" <dmagick@gmail.com>,
	"Pgsql performance" <pgsql-performance@postgresql.org>
References: <cone.1144953939.190214.95544.5001@35st-server.simplicato.com>
In-Reply-To: <cone.1144953939.190214.95544.5001@35st-server.simplicato.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200604141400.38982.mcousin@sigma.fr>

I hope I'm not going to say stupid things, but here's what i know (or i think 
i know :) ) about bacula + postgresql

If I remember correctly (I allready discussed this with Kern Sibbald a while 
ago), bacula does each insert in its own transaction : that's how the program 
is done, and of course it works ok with mysql and MyIsam tables, as mysql 
doesn't have transactions with myisam...

So, you'll probably end up being slowed down by WAL fsyncs ... and you won't 
have a lot of solutions. Maybe you should start with trying to set fsync=no 
as a test to confirm that (you should have a lot of iowaits right now if you 
haven't disabled fsync).

For now, I only could get good performance with bacula and postgresql when 
disabling fsync...


On Thursday 13 April 2006 20:45, Francisco Reyes wrote:
> Chris writes:
> > If you can, use copy instead:
> > http://www.postgresql.org/docs/8.1/interactive/sql-copy.html
>
> I am familiar with copy.
> Can't use it in this scenario.
>
> The data is coming from a program called Bacula (Backup server).
> It is not static data.
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster