Date: Wed, 17 Aug 2005 10:01:28 +1000
From: David Chinner <dgc@sgi.com>
To: djani22@dynamicweb.hu
Cc: linux-xfs@oss.sgi.com
Subject: Re: XFS-log +performance question
Message-ID: <20050817000128.GA1798768@melbourne.sgi.com>
References: <001001c59bfb$1fe7aae0$0400a8c0@LocalHost> <42F74B6D.8060002@gmx.net> <00d801c59c20$e0354080$0400a8c0@LocalHost> <42F798AF.5080505@gmx.net> <022d01c5a29a$bf3df860$0400a8c0@LocalHost>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <022d01c5a29a$bf3df860$0400a8c0@LocalHost>
User-Agent: Mutt/1.4.1i
Sender: linux-xfs-bounce@oss.sgi.com
Errors-to: linux-xfs-bounce@oss.sgi.com
Precedence: bulk
Content-Length: 1233
Lines: 45

On Tue, Aug 16, 2005 at 09:43:07PM +0200, djani22@dynamicweb.hu wrote:
> Hello list,
> 
> I have a little performance problem on my 8TB system.

......

> I searching this problems source for a long time, and I found one reason:
> 
> The XFS's log is always on node 3 (on the third 32k chunk).
> Am I right?

The internal log is located in the middle (by block number) of the
filesystem by default.  Where this is physically located on disk
depends on your device layout.

If you really have a RAID0 stripe, then the log will be striped
across all the disks in the RAID0 stripe.

> It is possible to get out the log to an alternative free and fast device
> without re-"format" the whole filesystem?

Ob-FAQ: Google is your friend.

See the link below for details. However, I'll repeat the BFW here:

PRACTICE THIS ON A DEVICE YOU DO NOT CARE ABOUT FIRST!!!

http://oss.sgi.com/archives/linux-xfs/2003-11/msg00171.html

> Additionally the node-2 and node-3 reads about 4x - 5x more blocks than the
> 1 and 4, but I don't know why.
> If somebody has an idea, let me know! :-)

We never read from the log, so I doubt the log is responsible
for this.

Cheers,

Dave.
-- 
Dave Chinner
R&D Software Enginner
SGI Australian Software Group