1.2.9 unstable ? (was [OpenAFS] inaccessibble volume - please help)

jarausch@igpm.rwth-aachen.de jarausch@igpm.rwth-aachen.de
Thu, 08 May 2003 09:08:57 +0200 (CEST)


I changed two things. First I upgraded from Linux kernel
2.4.21-rc1-ac2 to 2.4.21-rc1-ac4 (there was at least an
IDE hang at shutdown in the previous version)

and I think more importantly I stepped back to OpenAFS 1.2.8
and all my problems went away immediately.

Has anybody experienced similar problems with 1.2.9 ?

Thanks for your help,
Helmut.

On Wed, 7 May 2003 Derrick J Brashear <shadow@dementia.org> wrote:
> On Wed, 7 May 2003 jarausch@igpm.rwth-aachen.de wrote:
> 
> > Hi all
> > please bear in mind that I not an experienced afs admin.
> > Indeed, it's my first (tiny) afs installation with only a single
> > server. This holds several volumes on partitions /vicepa .. /vicepc
> > All except one volume are still accessible. I think I have seen an
> > error message indicating something like disk full or full volume
> > during a rsync transfer sunday night. But 'vos partinfo localhost'
> > indicates lots of free space on each partition. Since the failing
> > volume cannot be accessed anymore I cannot say if it's full.
> > The BosLog for this sunday night shows a lot of messages like
> > fs:vol exited on signal 15
> > fs:salv exited with code 0
> > fs:file exited with code 1
> 
> kill all the fileserver processes on the machine with kill -9.
> 
> the pthread problem of processes not dying completely is still there, in
> different form.
> 
> > Could not fetch the list of volumes from the server
> > Possible communication failure
> > Error in vos listvol command.
> > Possible communication failure
> 
> no volserver, no communication.
> 
> > This communication errors looks strange to me.
> > All volumes are on localhost and 'bos status'
> > doesn't show any unusual status.
> 
> it doesn't show proc starts ever increasing?
> 
> > What can I do to recover such a volume and what
> > might have been the cause of the failure.
> 
> it's not clear any recovery is needed, yet.

-- 
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany