mismatching vinum configurations

[Marcel Moolenaar]

On Sun, Jul 20, 2003 at 08:27:00PM +0200, Joan Picanyol i Puig wrote:
> Hi,
> 
> I had a power failure, and the on-disk configuration for vinum went 
> bizarre. The logs read from disks are at http://biaix.org/pk/debug/ 
> (log.$DEVICE files). The logs in da0 (barracuda) are the ones obviously 
> wrong, I'm pretty sure the others are ok. Is this a 'virtually' dead 
> drive? Can I force vinum to use the other's drive configuration? What's 
> the less traumatic way to recover the data?

I had the same problem twice. Also a power failure. There was no way
I could fix it without trauma. I had to do the following (and I'm
not joking about the reboots):

WARNING: do not follow the steps below if you're not entirely sure
what you're doing unless you already have nothing to lose...

1. Nuke the complete configuration, but make sure you know exactly
   what it was.
2. Reconfigure vinum from scratch in exactly the same way it was
   before.
3. Reboot the machine. I got kernel panics trying to recover my
   vinum volumes and that only made matters worse (it always can).
4a.For mirrored volumes you should be able to mark one drive as up,
   and mark the other(s) as empty (ie manually set the state). This
   forces a complete resync. Let it finish and reboot the machine
   afterwards.
4b.For RAID5 volumes you manually have to mark all disks as up and
   rebuild the parity. After that it's probably good to reboot the
   machine avoid kernel panics. Reboot twice if you're paranoid.
   Verify that the parity is ok!
5. fsck(8).
6. Drink lot's of alcoholic beverages to recover yourself after you
   verified you recovered vinum.
7. Reboot yourself.

FYI,

-- 
 Marcel Moolenaar	  USPA: A-39004		 marcel at xcllnt.net