fsck and raidtools

[Lennart Sorensen]

On Sun, Sep 03, 2006 at 07:51:59PM -0500, Sy Ali wrote:
> fsck seems to go mental when I try to use it on an md device.  Is this 
> normal?
> 
> # fsck /dev/md0
> fsck 1.39 (29-May-2006)
> e2fsck 1.39 (29-May-2006)
> The filesystem size (according to the superblock) is 17101184 blocks
> The physical size of the device is 17101168 blocks
> Either the superblock or the partition table is likely to be corrupt!
> Abort<y>?
> 
> Going ahead with the repairs seems to do nothing.  This error will
> always reappear.
> 
> What's going on?  How can I check an md device if fsck always acts like 
> this?
> 
> It sounds reasonable that I would have to check the (in this case) two
> source drives individually.  Is that right?

It looks more to me like someone did and mkfs on the device, then
decided to make it a raid device, and the raid mirrored that original
formatting (minus the raid superblock) onto both devices.  This means
part of the filesystem was overwritten of course by the raid superblock.
One must always make a new filesystem after creating a raid, even if it
looks like you can actually mount the old filesystem that was on one of
the devices before hand.

So in short, I believe someone made a mistake while setting up the raid
and that the drive now has a slightly corrupted filesystem on it and
should be re-mkfs'd to fix it.  I have certainly never had fsck fail on
an md device before.

--
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wini/Mailing_lists