[OpenAFS] strange error in directory / volume

Derrick J Brashear shadow@dementia.org
Mon, 2 Dec 2002 12:22:21 -0500 (EST)


(please reply to the list and not to me)

On Mon, 2 Dec 2002, Jack Neely wrote:

> On Sun, Dec 01, 2002 at 01:05:20PM +0100, Rubino Geiß wrote:
> > We do have a strange error:
> > 
> > ll %gconf-xml-backend.lock/
> > ls: %gconf-xml-backend.lock/ior: No such file or directory
> > 
> > obviously something goes wrong with "ls".  if we use rm -rf to delete
> > the whole directory. the error is the same "No such file or directory /
> > Cannot delete nonempty dir"
> 
> I have this same issue.  Its cropped up for at least three users here, 
> including myself.  We are using Red Hat Linux 7.3/8.0 with OpenAFS 1.2.7.
> My back end servers are Transarc AFS 3.6 on Solaris.  Looks like some
> strange bug...
> 
> Currently, my only work around is to move the directories containing the
> foobar'd GConf lock files so I the user can log in again.  (Then pray it
> doesn't happen again.)

i assume you can't boil it down into a simpler test case than using GConf? 

if not, i suggest:
-use a recent tcpdump and watch the network to see exactly what
operations are sent (tcpdump -vv -s 1500 port 7000 and host yourhost 
 should be sufficient)
-make sure fstrace is configured usefully (afszcm.cat installed in
 /usr/vice/etc/C or wherever you've compiled afs to expect it)
-enable fstrace on the client (fstrace setset cm -active) , make the
 problem happen, fstrace dump cm > somefile
-send the packet trace and the fstrace dump to openafs-bugs@openafs.org