Message-ID: <4C5310C4.1040906@emolecules.com>
Date: Fri, 30 Jul 2010 10:49:56 -0700
From: Craig James <craig_james@emolecules.com>
User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US;
	rv:1.9.1.11) Gecko/20100711 Thunderbird/3.0.6
MIME-Version: 1.0
To: Kevin Grittner <Kevin.Grittner@wicourts.gov>
CC: pgsql-admin@postgresql.org
Subject: Re: Connection pooling for a mixture of lightweight		 and
	heavyweight jobs?
References: <4C50ABA5.5010408@emolecules.com>
	<4C52B0240200002500033F10@gw.wicourts.gov>
	<4C530758.7050804@emolecules.com>
	<4C52C7790200002500033F1C@gw.wicourts.gov>
In-Reply-To: <4C52C7790200002500033F1C@gw.wicourts.gov>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 7/30/10 10:37 AM, Kevin Grittner wrote:
> Craig James<craig_james@emolecules.com>  wrote:
>
>>> Well, the "if it ain't broke, don't fix it" rule might come into
>>> play here.
>>
>> I should have given one more detail here: We've been the victim
>> of persistent "CPU spikes" that were discussed extensively in
>> postgres-performance.  Tom suggested upgrading to 8.4.4, but that
>> can't happen for a couple more months (we're working on it).
>>
>>
> http://archives.postgresql.org/pgsql-performance/2010-04/msg00071.php
>
> Ah, I hadn't connected that thread with this.  After rereading that
> thread with the information from this thread in mind, I think what
> you describe on the other thread could well be the "thundering herd"
> problem.  Some form of connection pooling could well help.
>
> BTW, I hope you've updated to the latest 8.3.x by now.  If not, you
> should expedite that.

Yes, I updated to 8.3.10, partly to see if it would solve this problem.

I'm not clear on how connection pooling would help this problem.  I would have 100 lightweight backends, whether they were pooled or not, always sitting around.  Or are you suggesting that I not use Apache::DBI to maintain persistent connections, and instead rely on the connection pooler to provide fast connect/disconnect from Postgres?

Thanks,
Craig