On Fri, 27 Apr 2012 16:04:22 -0400 likewhoa <likewhoa@xxxxxxxxxxxxxxxx> wrote:
> I have a strange issue on my raid10 8x400GB array. I cannot assemble my
> array anymore and it's gotten to a point that I don't know what to do to
> recover my data. I was hoping I could get some advice on it. Below is
> some info which I hope will help troubleshoot this issue.
>
> > mdadm -Esv
> ARRAY /dev/md/1 metadata=1.0 num-devices=0
> UUID=828ed03d:0c28afda:4a636e88:7b29ec9f name=Darkside:1
> spares=7
> devices=/dev/sdg3,/dev/sdf3,/dev/sde3,/dev/sdd3,/dev/sdc3,/dev/sda3,/dev/sdb3
> ARRAY /dev/md/1 level=raid10 metadata=1.0 num-devices=8
> UUID=828ed03d:0c28afda:4a636e88:7b29ec9f name=Darkside:1
> devices=8:131
I'm afraid you've been bitten by a rather nasty bug which is present in 3.3
and got back ported to some -stable kernel. The fix has been submitted and
should be appearing in -stable kernels soon (maybe already).
The effect of this bug is to remove key data from the metadata. It makes the
devices look like a spare for an array of unknown level with no devices and
no chunk size.
It seems that you have one device which wasn't corrupted: 8:131 which should
be /dev/sdi3. The others are messed up.
What you will need to do is re-create the array. You can look at /dev/sdi3
to see what the chunksize and layout is:
mdadm --examine /dev/sdi3
and then run a command like:
mdadm --stop /dev/md1
mdadm --create /dev/md1 --metadata=1.0 -l10 -n8 --chunk=XXX --layout=YY \
--assume-clean \
/dev/sdg3 /dev/sda3 /dev/sdi3 /dev/sdb3 /dev/sdf3 /dev/sde3 /dev/sdc3 /dev/sdh3
(probably have some device names wrong)
Note that the order of the devices is *very* important, and that is
information that has been lost. However some of it can be recovered.
If this is an 'n2' or 'near=2' array, which is the default, then adjacent
devices will be identical. You can find identical pairs with:
for i in a b c d e f g h i
do for j in a b c d e f g h i
do if [ $i = $j ] ; then continue fi
if cmp -s -n 32768 /dev/sd${i}3 /dev/sd${j}3
then echo /dev/sd${i}3 and /dev/sd${j}3 seem to match
fi
done
done
That should help you find pairs. Then you have at most 16 possible orderings
of those 4 pairs to test.
Note that the "--create" command doesn't touch the data. It just writes new
metadata and assembles the array. Then you can run "fsck -n" to check if the
array looks good.
Good luck, and please ask if anything here isn't clear.
NeilBrown
Attachment:
signature.asc
Description: PGP signature
