Date: Sat, 3 Sep 2005 22:43:20 +0300
From: Iustin Pop <iusty@k1024.org>
To: A JM <vbtalent@gmail.com>
Cc: evilninja <evilninja@gmx.net>, linux-xfs@oss.sgi.com
Subject: Re: XFS - hard drive dying
Message-ID: <20050903194320.GA6857@saytrin.hq.k1024.org>
Mail-Followup-To: A JM <vbtalent@gmail.com>,
	evilninja <evilninja@gmx.net>, linux-xfs@oss.sgi.com
References: <a0f05e8905090215183e30f40c@mail.gmail.com> <a0f05e8905090303584075a22e@mail.gmail.com> <4319B8DD.7080200@gmx.net> <a0f05e8905090311012f7b018e@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <a0f05e8905090311012f7b018e@mail.gmail.com>
User-Agent: Mutt/1.5.10i
Sender: linux-xfs-bounce@oss.sgi.com
Errors-to: linux-xfs-bounce@oss.sgi.com
Precedence: bulk
Content-Length: 2357
Lines: 57

On Sat, Sep 03, 2005 at 02:01:29PM -0400, A JM wrote:
>  Just so that I understand and know that I have creted the backup device 
> correctly please tell me if the following steps were correct.
>  1) I deleted all partitions on the destination device.
Good.
>  2) I dd_rescued to the new device using the following 'dd_rescue /dev/hdd 
> /dev/hdb' (hdd being the bad drive and hdb being the good drive)
Good.
> 3) I ran xfs_repair /dev/hdb.
This is wrong. /dev/hdb is a partioned disk, not a XFS filesystem.
>   I just looked at the dev/hdb using fdisk and this is what it shows.
 
> Device Boot Start End Blocks Id System
> /dev/hdb1 1 24792 199141708+ 8e Linux LVM 
>   So, it appears to have copied the information over correctly because the 
> entire drive was used in the LVM.
Yes, but this means you can't xfs_repair /dev/hdb or /dev/hdb1

> Should I have used /dev/hdb1 when trying xfs_repair assuming it's an xfs 
> file system?
No, /dev/hdb1 is a PV not a filesystem.

What I understand from your emails is this:
 - you had a LVM setup, with /dev/hdd one of your physical volumes
 - /dev/hdd failed and you managed to copy parts of it in /dev/hdb
 - you are trying xfs_repair /dev/hdb or /dev/hdb1

If the first two points are correct, then you are wrong in the third
step. /dev/hdb1 in not in any kind of way an XFS filesystem. It's a PV,
and you won't be able to xfs_repair it, because it probably starts with LVM
metadata and the offsets are thus wrong in the file - even if you manage
to reach a valid superblock.

What you need to do, if I understood correctly your situation:

1. try to re-activate your VG - VGforMyth. It is important that you do
this with the failed harddrive (/dev/hdd) not in the system. First, do:
# pvscan
this should show you all the PV of the VGforMyth, with /dev/hdb1 being
now in the place of /dev/hdd1. Then do a vgchange -a y VGforMyth.

If this is not successfull, please post the output of these commands,
the output of pvdisplay /dev/hdb1 and the contents of the /etc/lvm
directory (if it's not too big). Especially the /etc/lvm/archive/* right
before your harddrive failed.

2. after you have activated correctly the VG, then it is time to
xfs_repair the correct logical volume:
# xfs_repair -n /dev/VGforMyth/video

This should work, if the LVM configuration is sane.

Hope this helps!

Regards,
Iustin Pop