Re: Occasional Slow Commit - Mailing list pgsql-performance

From David Rees
Subject Re: Occasional Slow Commit
Date
Msg-id 72dbd3150811060747u7f3c2aadv3c6658d285d74455@mail.gmail.com
Whole thread Raw
In response to Re: Occasional Slow Commit  (Peter Schuller <peter.schuller@infidyne.com>)
Responses Re: Occasional Slow Commit
List pgsql-performance
On Thu, Nov 6, 2008 at 2:21 AM, Peter Schuller
<peter.schuller@infidyne.com> wrote:
>> I also found that my write cache was set to WriteThrough instead of
>> WriteBack, defeating the purpose of having a BBU and that my secondary
>> server apparently doesn't have a BBU on it. :-(
>
> Note also that several RAID controllers will periodically drop the
> write-back mode during battery capacity tests. If you care about
> consistently/deterministically having full performance (with
> white-backed battery protected caching), you probably want to confirm
> your controller behavior here.
>
> (I've seen this on at least LSI based controllers in Dell 2950:s, and
> also on some 3wares.)

I can confirm that this is the case by reviewing the logs stored on
the MegaRAID controller that have a BBU and had WriteBack configured.
The controller also lets you know (using the MegaCli utility) what
setting is configured, and what setting is in effect.

In the case of the machines without a BBU on them, they are configured
to be in WriteBack, but are actually running in WriteThrough.

-Dave

pgsql-performance by date:

Previous
From: "Helio Campos Mello de Andrade"
Date:
Subject: Re: PostgreSQL OR performance
Next
From: "Scott Marlowe"
Date:
Subject: Re: Occasional Slow Commit