Re: High Load on Postgres 7.4.16 Server

From: Jeff Frost <jeff(at)frostconsultingllc(dot)com>
To: John Allgood <jallgood(at)the-allgoods(dot)net>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: High Load on Postgres 7.4.16 Server
Date: 2007-04-05 19:23:33
Message-ID: Pine.LNX.4.64.0704051218470.32187@glacier.frostconsultingllc.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Thu, 5 Apr 2007, John Allgood wrote:

> Hello All
>
> I sent this message to the admin list and it never got through so I
> am trying the performance list.
> We moved our application to a new machine last night. It is a Dell
> PowerEdge 6950 2X Dual Core. AMD Opteron 8214 2.2Ghz. 8GB Memory. The
> machine is running Redhat AS 4 Upd 4 and Redhat Cluster Suite. The SAN is an
> EMC SAS connected via fibre. We are using Postgres 7.4.16. We have recently
> had some major hardware issues and replaced the hardware with brand new Dell
> equipment. We expected a major performance increase over the previous being
> the old equipment was nearly three years old
> I will try and explain how things are configured. We have 10
> separate postmasters running 5 on each node. Each of the postmasters is a
> single instance of each database. Each database is separated by division and
> also we have them separate so we can restart an postmaster with needing to
> restart all databases My largest database is about 7 GB. And the others run
> anywhere from 100MB - 1.8GB.
> The other configuration was RHEL3 and Postgres 7.4.13 and Redhat
> Cluster Suite. The application seemed to run much faster on the older
> equipment.
> My thoughts on the issues are that I could be something with the OS
> tuning. Here is what my kernel.shmmax, kernel.shmall = 1073741824. Is there
> something else that I could tune in the OS. My max_connections=35 and shared
> buffers=8192 for my largest database.

John,

Was the SAN connected to the previous machine or is it also a new addition
with the Dell hardware? We had a fairly recent post regarding a similar
upgrade in which the SAN ended up being the problem, so the first thing I
would do is test the SAN with bonnie-++ and/or move your application to use a
local disk and test again. With 8GB of RAM, I'd probably set the
shared_buffers to at least 50000...If I remember correctly, this was the most
you could set it to on 7.4.x and continue benefitting from it. I'd strongly
encourage you to upgrade to at least 8.1.8 (and possibly 8.2.3) if you can, as
it has much better shared memory management. You might also want to double
check your effective_cache_size and random_page_cost to see if they are set to
reasonable values. Did you just copy the old postgresql.conf over?

This is the beginning of the thread I mentioned above:

http://archives.postgresql.org/pgsql-performance/2007-03/msg00104.php

--
Jeff Frost, Owner <jeff(at)frostconsultingllc(dot)com>
Frost Consulting, LLC http://www.frostconsultingllc.com/
Phone: 650-780-7908 FAX: 650-649-1954

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message C. Bergström 2007-04-05 19:28:23 Re: High Load on Postgres 7.4.16 Server
Previous Message John Allgood 2007-04-05 19:06:14 High Load on Postgres 7.4.16 Server