Re: Brocken Raid & LUKS

[Phil Turmel <philip@turmel.org>]

On 02/21/2013 12:47 PM, Stone wrote:
> Am 21.02.2013 18:36, schrieb Phil Turmel:

>> But it does give you the locations of the backup superblocks.  Use these
>> numbers, starting with 32768, as "xxxx" in:
>>
>> fsck.ext4 -n -b xxxx /dev
>>
>> Once you give it a superblock that hasn't been corrupted, it should be
>> able to check the rest of the filesystem.  There will be damage near the
>> beginning, and probably more damage where you had to put zeros.
>>
>> If it looks like that, do it again without "-n" to actually fix it.
>>
>> Phil
>>
> ok. i think i dont understand you not complete.
> i restore now the superblock with --> fsck.ext4 -bv 4096000
> /dev/mapper/md2_nas

No!  You must keep using "-n" until you have seen a mostly-clean report!
 We don't know yet that the chunk size is right.

Leaving off "-n" will simultaneously fix the superblock (and all other
backup copies) and continue to fix the rest of the filesystem.  You
mustn't do that with a wrong chunk size--it will damage much more.

> when this is done i make full filesystemcheck --> fsck.ext4 /dev/md2_nas
> and answer the quest questions.

> right?

No.

Just do the "fsck -n -b xxxx" combinations until you find a good
superblock.  That report will also show if there are many other errors.
 Expect scattered damage in the region < 384MB due to the wrong data
offset.  After that, there should only be errors where the new zeros
are, and maybe a few scattered errors from the original crash.  If there
many more errors, you have the wrong chunk size.

Phil