[linux-lvm] thinpool undeleteable after server crash

Zdenek Kabelac

2014-02-10 16:21:18 UTC

Post by Oliver Rath
Hi list,
Im using here lvm 2.02.104 with kernel 3.11 and today the server
crashed, so I had to power off and reboot the machin (dont know why at
the mom). After this, the used thinpool-device got an unrecoverable
error (thin_repair didnt help) so i tried to remove the whole vg
including the thinpool and its depending devices, which didnt work.
My only working solution was to overwrite the the first 30MB or the
beginning of the used pv with zeroes (dd if=/dev/zero of=/dev/sda1),
then to reboot and NOW to recreate the whole thing.
Is it possible to force recreate of a thinpool device when it is
crashing? What can I do to avoid unrecoverable errors on thinpools? Is
there a better possibility for recreation of pv/vg/thinpools?

Developer are quite interested in any thin pool corruptions you could have.

There is of course 'final way' to remove thin pool -
'vgcfgbackup' your metadata - removed offending volumes
(thinpool + thin volumes) in text editor
and 'vgcfgrestore'.

But - thin recovery tools are there to fix it - and if they fail,
it would be good to get such problematic metadata for analysis,
before they are removed.

Zdenek