0

I am completely lost :( I got software RAID10 with 4 drives, one of them failed yesterday, right before or after the failure system crashed and restarted. It started data rebalancing and completed it successfully, however it still can't mount raid10 partition due to file system errors. When tryin to do: mount /dev/md2 /mnt/

It gives an error in syslog: EXT4-fs (md2): ext4_check_descriptors: Checksum for group 1920 failed (12014!=34594)

Naturally I started fsck on that partition. It ran for almost a day, and pretty much stuck fixing multiple-claimed blocks. Seems like every singly inode has thousands of them.

Is it possible to mount corrupted ext4 to get whatever data out of it that I can get? If not, will letting fsck complete the procedure(could take weeks) will help anyhow?

Alex N
  • 782
  • 3
  • 10
  • 24

2 Answers2

1

You do have a backup, right ?

You may try to mount the filesystem read-only:
mount -o ro,noload /dev/md2 /mnt

But it is possible that will not work. In which case, you would need to let fsck do its work, or go back to a recent backup.

Do you use software or hardware RAID ?

Urgoll
  • 681
  • 3
  • 6
  • it's a software raid, I only have a backup from last month ;( read only mount actually did work, but half of the folders are missing... – Alex N Aug 31 '10 at 17:23
  • You could try forcing a scan/replair on the sotfware raid: echo "repair" > /sys/block/md2/md/sync_action Do not forget to umount first, then run fsck again. – Urgoll Aug 31 '10 at 17:34
0

Depending what data you want, and I know this isn't the answer you're looking for (or probably either the best answer), you can grep through a device if you specify the -a flag. I've used it to recover deleted text files in the past. Also, you could try a dd onto a spare disk and see if restores anything useful.

James L
  • 6,025
  • 1
  • 22
  • 26