Message-Id: <4.3.2.7.2.20020521085022.03905aa8@pop.xs4all.nl>
Date: Tue, 21 May 2002 09:24:05 +0200
To: Sidik Isani <lksi@cfht.hawaii.edu>, Steve Lord <lord@sgi.com>
From: Seth Mos <knuffie@xs4all.nl>
Subject: Re: "Corruption of in-memory data"
Cc: Sidik Isani <lksi@cfht.hawaii.edu>, linux-xfs@oss.sgi.com
In-Reply-To: <20020520102853.C18897@cfht.hawaii.edu>
References: <1021923461.4832.335.camel@jen.americas.sgi.com>
 <20020520090515.B18897@cfht.hawaii.edu>
 <1021923461.4832.335.camel@jen.americas.sgi.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

At 10:28 20-5-2002 -1000, Sidik Isani wrote:

>   It's a software raid5, and only one of 6 disks failed.  To XFS,
>   the device should have been completely functional.

That's the whole idea :-)

> > looks something like that. Destruction in these areas also appears
> > pretty drastic.
>
>   Oh, there's no worry about the contents of this filesystem.  I was
>   actually trying to do the benchmarks on performance that Seth
>   mentioned had not been done.  If I get them done, I'll post the
>   results here.  I'd just really like to understand what happened.
>   If raid5 is to blame here, there isn't much point in using it!
>   Any suggestions on the best way to narrow it down?
>
>   Maybe it is worth starting over, with the bad disk still in there,
>   to see if it happens again.  The sequence was:

I installed my software raid5 with xfs in degraded mode.
It consisted of a 3 disk array with one disk as a failed (which still 
contained data).
Formatted the degraded array with XFS.
Copied the data to the degraded raid5 array.
Added the old data disk into the raid 5 array and started reconstruction.

This worked for me.
It might be that the error during resyncing upset the raid/xfs/ide layer 
for reasons which can not be explained.

Cheers

--
Seth
It might just be your lucky day, if you only knew.