Re: Recover array after I panicked

["Patrik Dahlström" <risca@powerlamerz.org>]

On 04/23/2017 12:23 PM, Patrik Dahlström wrote:
> On 04/23/2017 12:16 PM, Andreas Klauer wrote:
>> On Sun, Apr 23, 2017 at 11:47:34AM +0200, Patrik Dahlström wrote:
>>> Is there any help you can offer?
>>
>> Is there any mdadm --examine output?
> At this point, it is incorrect. I've lost the output from the working
> raid too, unless it's located in any log in /var/log/.
> I have /etc/mdadm/mdadm.conf, but don't know if it's updated.
> 
>>
>> What was on the array? Regular filesystem, unencrypted, or LVM, LUKS, ...?
> Regular filesystem, unencrypted ext4.
> 
>>
>> If it's LUKS encrypted and you had RAID metadata at the end, yet 
>> mdadm --create'd new metadata at the start, that would likely have 
>> damaged your LUKS header beyond repair (and regular filesystems 
>> don't like it, either).
> No file system encryption.
> 
>>
>> If it's unencrypted data, as a last resort you can always go and find 
>> the header of a large file of known type... for example if you find 
>> a megapixel JPEG image and the first 512K of it are part of that then 
>> your chunksize would be 512K and then you can go looking for the 
>> next chunk on the other disks... and that should give you some notion 
>> of the RAID layout and offset.
> That's not a bad idea. Will hopefully narrow down my unknown variables.
Okay, I extracted parts of an mkv file and this is what I found out:
* playing 512 kb of data is OK
* playing 1024 kb of data will give me the following error (from mpv):
[mkv] Invalid EBML length at position 539473
[mkv] Corrupt file detected. Trying to resync starting from position
539473...

"position 539473" is at ~527 kb, which leads me to suspect that the
correct chunk size is 512 kb.

Any thoughts?

> 
>>
>> Regards
>> Andreas Klauer
>>