pg.ddx.io  pgsql-performance@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Jérôme BENOIS <benois@argia-engineering.fr>
To: Markus Schaber <schabi@logix-tt.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: High CPU Load
Date: Mon, 18 Sep 2006 16:44:05 +0200
Message-ID: <1158590646.5665.71.camel@localhost.localdomain> (raw)
In-Reply-To: <450A75DE.6050102@logix-tt.com>
References: <1158239339.5226.17.camel@localhost.localdomain>
	<450A75DE.6050102@logix-tt.com>

Hi Markus,

Le vendredi 15 septembre 2006 à 11:43 +0200, Markus Schaber a écrit :
> Hi, Jérôme,
> 
> Jérôme BENOIS wrote:
> 
> > max_connections = 512
> 
> Do you really have that much concurrent connections? Then you should
> think about getting a larger machine, probably.
> 
> You will definitely want to play with commit_delay and commit_siblings
> settings in that case, especially if you have write access.
> 
> > work_mem = 65536
> > effective_cache_size = 131072
> 
> hmm, 131072*8*1024 + 512*65536*1024 = 35433480192 - thats 33 Gig of
> Memory you assume here, not counting OS usage, and the fact that certain
> queries can use up a multiple of work_mem.

Now i Have 335 concurrent connections, i decreased work_mem parameter to
32768 and disabled Hyper Threading in BIOS. But my CPU load is still
very important.

Tomorrow morning i plan to add 2Giga RAM ... But I don't understand why
my database server worked good with previous version of postgres and
same queries ...

> Even on amachine that big, I'd be inclined to dedicate more memory to
> caching, and less to the backends, unless specific needs dictate it. You
> could try to use sqlrelay or pgpool to cut down the number of backends
> you need.
I used already database pool on my application and when i decrease
number of connection my application is more slow ;-(
> 
> > My Server is Dual Xeon 3.06GHz
> 
> For xeons, there were rumours about "context switch storms" which kill
> performance.
I disabled Hyper Threading.
> > with 2 Go RAM and good SCSI disks.
> 
> For 2 Gigs of ram, you should cut down the number of concurrent backends.
> 
> Does your machine go into swap?
No, 0 swap found and i cannot found pgsql_tmp files in $PG_DATA/base/...
> 
> Markus
-- 
Jérôme,

python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'sioneb@gnireenigne-aigra.rf'.split('@')])"

Attachments:

  signature.asc (application/pgp-signature, 190B) download

view thread (30+ messages)  latest in thread

Message-ID: <1158590646.5665.71.camel@localhost.localdomain>
Permalink:  ../1158590646.5665.71.camel@localhost.localdomain/
Also on:    postgresql.org/message-id/1158590646.5665.71.camel@localhost.localdomain

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-performance@postgresql.org
  Cc: benois@argia-engineering.fr, schabi@logix-tt.com
  Subject: Re: High CPU Load
  In-Reply-To: <1158590646.5665.71.camel@localhost.localdomain>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox