[PLUG] mdadm raid1 creation stomped on jfs superblock?

[Russell Senior]

In the process of recovering my rsync snapshots, I ended up with a
surplus of 200 gig disks.  My intention is to build a raid1 array to
store one of the snapshot trees with two extra disks.  I had hoped to
take an existing filesystem (and a dd'd copy) and transform it into a
raid1 array.  I did the "mdadm --create" [1] and mounted it read-only
with no hiccups.  However, when I boot, I get this message during the
fsck:

   The current drive is:  /dev/md1

   Superblock is corrupt and cannot be repaired
   since both primary and secondary copies are corrupt.

     CANNOT CONTINUE

   fsck failed.  Please repair manually.

I assume [??] this is because mdadm stomped on jfs somehow that jfs
doesn't like[2].  Do I have to re-mkfs and then rebuild the snapshot
tree from another backup or is there something simpler I can do?  I'd
ordinarily have done it in the Right Order[tm], but building the
snapshot trees is tedious.  Hell, just dd'ing disk to disk takes about
3 hours.  Clues?

Thanks!

[1] Like so:

    # mdadm --create /dev/md1 --auto=yes --level=1 --raid-devices=2 /dev/hdm1 /dev/hdp1

[2] Strace'ing the fsck.jfs, I see to looking at two locations (32768
    and 61440) and doing 4k reads immediately prior to the "is
    corrupt" message.



-- 
Russell Senior         ``I have nine fingers; you have ten.''
seniorr at aracnet.com