On Fri, 2005-12-16 at 06:45, Hendrik Naumann wrote:

Hi

I have a Linux-Software-Raid5 consisting of three partitions (hda2,
hde2 and hdf2). After a failure of hdf I replaced the disk and added
hdf2 into the raid again. During the resynchronization process hda
broke down. 

I am not completely shure how much of the synchronization
was already done, but I think it was quite at the end, because the 
gzipped dd-image of hdf2 and hde2 have very similar sizes. 

mdadm shows hde2 as a active, clean disk within the raid and hdf2 as a
clean spare disk. In the Software-Raid-Howto it was stated that
rewriting the superblocks (I suspect hdf2 to clean active) could be a
possibility to regain some data from the Raid. 

How is this possible with mdadm? Or is there some other tool for
manipulation the superblocks?

Many thanks in advance

Hendrik Naummann 

PS. Please add a CC hn75@gmx.de because up to now I am no subscriber 
to this list.

Since the recovery did not finish, I would do the followings:

mdadm --examine /dev/hde2 to display the raid5 configuration and note
the chunksize, parity layout, disks order,..
Recreate the array using mdadm -C with "missing" keyword
Mount the raid5 array read-only (hope that it will mount :)

--
Mike T.