Message-Id: <4.3.2.7.2.20040115084408.044a32e0@pop.xs4all.nl>
Date: Thu, 15 Jan 2004 08:48:00 +0100
To: Stephan L Jansen <jansen@astro.wisc.edu>, linux-xfs@oss.sgi.com
From: Seth Mos <knuffie@xs4all.nl>
Subject: Re: xfs_force_shutdown from line 1071 of xfs_trans.c
In-Reply-To: <A0366F44-46DF-11D8-82FD-003065F970BA@astro.wisc.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: linux-xfs-bounce@oss.sgi.com
Errors-to: linux-xfs-bounce@oss.sgi.com
Precedence: bulk
Content-Length: 1216
Lines: 36

At 16:18 14-1-2004 -0600, Stephan L Jansen wrote:

>Hi,
>
>I have a ~ 80GB parition (5 disk hardware RAID 0) on a two processor Dell 
>2650.  I
>have seen two xfs_force_shutdowns on it in the last few days.  The 
>filesystem is
>accessed largely via NFS by between 50 or 100 processes on remote 
>machines.  I'm
>currently running 2.4.20-28_36.rh8.0.atsmp which is XFS 1.3.0.  There are 
>no other
>logs in either syslog or dmesg.  Here's the error message:

If it's a Perc3/Di it's a adaptec (aacraid) controller. Basically it is 
crap piece of hardware that falls apart under load.

There is a workaround which you can found on the just launched 
http://linux.dell.com/

The scsi layer in the Perc3 gets stuck often which results in the 
controller firmware deadlocking. Upping a certain timeout in the driver to 
60 seconds will prevent this from throwing a IO error to the kernel but the 
box will stall for a seemingly long time.

The best workaround for the moment is using either software raid or 
purchasing a Perc3/DC or better yet a Perc4/DC.
Both are also faster and a lot faster respectively with regards to raid 5 
throughput.

Cheers

--
Seth
I don't make sense, I don't pretend to either. Questions?