[OpenAFS] UFS logging on Solaris - could this cause volumes to go offline and salvager to dump core

Douglas E. Engert deengert@anl.gov
Fri, 02 Nov 2007 09:25:34 -0500

Andrew Cobaugh wrote:
> I can't find any reference to UFS logging on Solaris past 2004 in the
> archives (unless I'm not looking hard enough). Is it safe to turn on
> logging on ufs under Solaris these days?

Don't thing anything has changed. Logging should be off. We used to
use UFS for vice, but have now switched to using the namei with ZFS.

A few years ago we ran into problems with fsck, as AFS required its
own fsck, because the AFS fileserver dinks around with UFS inodes.
But Sun change the superblock for logging and the AFS fsck could not
handle it, and did not do check the volume.

The long term solution is to use the namei fileservers.

> Reason I ask is I have one or two volumes that will randomly go
> offline, or fail to vos backup or move. Just yesterday my homedir
> volume went offline, and subsequent attempts to salvage the vice
> partition resulted in salvager dumping core. After the 3rd attempt it
> finally succeeded in reattaching the volume.
> This seems to occur regardless of what vice partition the volume is
> on, ruling out disk issues (and my disks are mirrored across separate
> controllers), and I don't see anything to indicate disk problems to
> that end.
> I am running Solaris 10 update 2 with OpenAFS 1.4.4. I wish I could
> attribute the volume offline-ing and salvager problem to logging
> (which I will turn off if that turns out to be a problem).
> I have the core dump and the output from SalvageLog from one run where
> it dumps core, and the final run where it succeeds here:
>     http://www.phys.psu.edu/~phalenor/salvager_problem/


  Douglas E. Engert  <DEEngert@anl.gov>
  Argonne National Laboratory
  9700 South Cass Avenue
  Argonne, Illinois  60439
  (630) 252-5444