Re: filesystem options for WAL

From: Greg Smith <gsmith(at)gregsmith(dot)com>
To: Jaime Casanova <jcasanov(at)systemguards(dot)com(dot)ec>
Cc: psql performance list <pgsql-performance(at)postgresql(dot)org>
Subject: Re: filesystem options for WAL
Date: 2008-07-06 15:18:24
Message-ID: Pine.GSO.4.64.0807061112570.15444@westnet.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Sun, 6 Jul 2008, Jaime Casanova wrote:

> Here http://www.westnet.com/~gsmith/content/postgresql/TuningPGWAL.htm I read:
> """
> Combining these two, an optimal fstab for the WAL might look like this:
>
> /dev/hda2 /var ext3 defaults,writeback,noatime 1 2
> """
> Is this info accurate?

Nah, that guy doesn't know what he's talking about. That article is
overdue for an overhaul.

> I also read on other document from the "technical documentation" that
> for partitions where you have the tables and indexes is better to have
> journaling and for partitions for the WAL is better to not have
> journalling...

The WAL is itself a sort of journal, and the way writes to it are done the
filesystem level journaling that ext3 provides doesn't buy you much beyond
additional overhead. Check out
http://www.commandprompt.com/blogs/joshua_drake/2008/04/is_that_performance_i_smell_ext2_vs_ext3_on_50_spindles_testing_for_postgresql/
for an extensive comparison of different options here, where you can see
that using ext2 instead can be much more efficient. The main downside of
ext2 is that you might get longer boot times from running fsck, but it
won't be any less reliable for database use though.

--
* Greg Smith gsmith(at)gregsmith(dot)com http://www.gregsmith.com Baltimore, MD

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Bill Moran 2008-07-07 11:15:46 Re: How much work_mem to configure...
Previous Message Scott Marlowe 2008-07-06 06:04:36 Re: How much work_mem to configure...