Received: from localhost (unknown [200.46.204.183]) by developer.postgresql.org (Postfix) with ESMTP id 471082E0074 for ; Mon, 16 Jun 2008 23:43:38 -0300 (ADT) Received: from developer.postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.204.183]) (amavisd-maia, port 10024) with ESMTP id 27731-01 for ; Mon, 16 Jun 2008 23:43:29 -0300 (ADT) X-Greylist: domain auto-whitelisted by SQLgrey-1.7.6 Received: from rv-out-0506.google.com (rv-out-0506.google.com [209.85.198.232]) by developer.postgresql.org (Postfix) with ESMTP id 1F1352E0073 for ; Mon, 16 Jun 2008 23:43:34 -0300 (ADT) Received: by rv-out-0506.google.com with SMTP id b25so6207848rvf.43 for ; Mon, 16 Jun 2008 19:43:31 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:cc:in-reply-to:mime-version:content-type :content-transfer-encoding:content-disposition:references; bh=mghwsh0V2TM9TlrosrZw0LCgbasynq3eQTPJ/I0/8iI=; b=LQFZCyzyNjdphl+bxkMCAZhXS++43g0yukmv/6z+DaWHZxsCWR8KgftPFGdih/Z8BO cl7l+jdcGjRV5EhRW19ifq2O3QPXyZTKWiI9khKUZXobmxTpIj7uNiXZOKxJCz2jJtsG tVI5JoayNXUQAORQVL0Z5mkMYjGq633fcMFq8= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=BvT565UWrwRHMHu+rhJMJYZjRpQeULZ+UnRTDx4sb2sBKXv0yR+Bd4PGE2AVemrKGm iuWM3kWfCW3WLTiPo5kFpyLQWtaHjPJjczTtL+OngvKlUBki+/KCeYbVn+xF01LedNBI I4WiarwELmFeY/Pl6ToK8PgJr30rRHmtX7JwU= Received: by 10.114.179.1 with SMTP id b1mr7221479waf.108.1213670610654; Mon, 16 Jun 2008 19:43:30 -0700 (PDT) Received: by 10.114.130.17 with HTTP; Mon, 16 Jun 2008 19:43:30 -0700 (PDT) Message-ID: Date: Mon, 16 Jun 2008 20:43:30 -0600 From: "Scott Marlowe" To: "Andreas Philipp" Subject: Re: Advice on running two database clusters on one server machine Cc: pgsql-admin@postgresql.org In-Reply-To: <72fa1b080806151011h357edc5bm92af30e08aeefd61@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <72fa1b080806151011h357edc5bm92af30e08aeefd61@mail.gmail.com> X-Virus-Scanned: Maia Mailguard 1.0.1 X-Archive-Number: 200806/130 X-Sequence-Number: 29468 On Sun, Jun 15, 2008 at 11:11 AM, Andreas Philipp wrote: > Hi all, > > We are implementing a hospital information system and a human > resources/payroll processing system on two identical dedicated servers with > two Xeon Quad Core processors and 32 GB RAM each, both servers being attached > via FC to a SAN, and both applications running on PostgreSQL 8.3 / CentOS 51. > > We are wondering about the advisability to distribute the databases between > the two server machines, both machines acting as active production systems > for one application each, and as warm standby servers for the other, using > WAL shipping to a second database cluster running on another port on each of > the two server machines. > > What would be the performance cost of doing so, rather than running all > databases on one database cluster on one machine, and using the second > machine as a warm standby server for all databases of the two applications? Well, when both machines are working, your performance would be better on two machines than on one. But after a failover, the warm standby will be running two instances of postgresql, and that's sub-optimal. > What other considerations should we take into account? We have no prior > experience with PostgeSQL administration, having run our previous systems on > Windows Servers and MS SQL Server. Also, worry about the possibility of switching the wrong databases when things are going wrong, etc with this type of setup. However, the biggest concern is your FC-SAN setup. Some are very fast. Some are quite pokey, and once you buy them, the manufacturer will just laugh and point their fingers at linux, postgresql, or the phase of the moon as the cause of your performance problem and do nothing more. If you haven't tested your FC-SAN setup with both servers running comlex benchmarks (multiple bonnie++ instances, lots of dds, etc...) and proven that it's fast, don't expect it to be.