[DRBD-user] Diskless / unclean metadata

[olc]

Hi -

On 24/02/2014 17:52, Lars Ellenberg wrote:
>> My primary node crashed today. I have had to reboot it but now it
>> reports unclean metadata. :(
>
> How about a simple "drbdadm adjust all"?
>
> If you insist on doing the steps manually,
> do a "drbdadm -d adjust all".
>
> To "up" a drbd resource with 8.4
> there are more steps involved than there were with 8.3...
> Which steps? See above.

Thanks Lars. I could not make "drbdadm adjust all" to work (no usable 
activity log). As I'm just concerned with getting back the data, I have 
tested to modify the uuid's thanks to dump-md/create-md/restore-md, 
following the guide at http://www.drbd.org/users-guide/s-gi.html

Initial situation:
uuid {
     0x3AD1EEC6E5D55491; 0x8F95E078DD5699B8; 0x67ACDDB56C762980; 
0x67ABDDB56C762981;
     flags 0x00000013;
}

Modified to:
uuid {
     0x8F95E078DD5699B8; 0x0000000000000000; 0x67ACDDB56C762980; 
0x67ABDDB56C762981;
     flags 0x00000013;
}

The drbd resource was then happy to start as primary. Unfortunately, I 
haven't been able to mount the ext4 partition: "wrong fs type, bad 
option, bad superblock on /dev/drbd0". I have tried many things (working 
on dd'ed device) but nothing doing: seems to be mission impossible. I 
have just been able to recover lots of file with photorec but this does 
not help so much (there is 50 billions files on the FS, no way to 
retrieve which is which without knowing the path/filename). Anyway, we 
are out of the drbd topic here: this is related to ext4.

BTW, in such situation when you don't care on the drbd's metadata, I 
think that you can just get rid of the drbd stack and mount the device 
directly as is. Knowing that could have saved lots of time here.

Regards,
-- 
Olivier