[Lustre-discuss] bad superblocks

Johann Lombardi johann.lombardi at oracle.com
Fri Oct 1 05:35:48 PDT 2010


On Thu, Sep 30, 2010 at 04:07:01PM -0400, Michael Di Domenico wrote:
> Had a hard power failure this morning on the whole building.  Now when
> I attempt to mount the metadata partition I get
> 
> VFS: Can't find ldiskfs filesystem on dev md22.
> LustreError: 6570:0:(obd_mount.c:1278:server_kernel_mount()) premount
> /dev/md22:0x0 ldiskfs failed: -22, ldiskfs2 failed: -19
> LustreError: 6570:0:(obd_mount.c:1592:server_fill_super()) Unable to
> mount device /dev/md22: -22
> LustreError: 6570:0:(obd_mount.c:1997:lustre_fill_super()) Unable to mount (-22)
> 
> This feels a bit like an uh-oh moment.  Any suggestions?

I would first check that the MD array is sane. Last time someone reported
a similar issue, it was a problem with the software raid configuration
(somehow, two disks got inverted in the array).
Then, if the superblock is still reported to be corrupted, you can try to
run e2fsck with one of the backup superblocks (e.g. with -B 4096 and -b 32768).

> e2fsck the partition fails, complains of a bad superblock.  I tried to
> dump out the superblocks, but that failed also, (though I'm not sure i
> did it right)

Ah, so you tried something like "dumpe2fs -o superblock=32768 $dev"
or "dumpe2fs -o superblock=98304 $dev" already?

Cheers,
Johann



More information about the lustre-discuss mailing list