[nfsv4] [nfsv4] server-cluster-locking-api: cthon failure

[Bryce Harrington]

On Wed, Jun 28, 2006 at 02:33:38PM -0400, Fredric Isaman wrote:
> 
> 
> On Wed, 28 Jun 2006, Bryce Harrington wrote:
> 
> > Ahh, okay thanks.  I don't mind tracking down the problems but of course
> > would prefer doing it for _un_known issues.  ;-)
> >
> > Before I work on these next two, could you see if these are also already
> > known, and/or if it's worth it to dig into them further?
> >
> > 1.  PyNFS - http://linux-nfs.org/pipermail/nfsv4/2006-June/004466.html
> >
> >     Looking through the code, it appears that there is a dependency,
> >     MKFILE, which fails and appears to cause a number of OMITS (and,
> >     possibly, also causing the increased number of FAILS we see).
> 
> Looking at the log at
> http://developer.osdl.org/dev/nfsv4/results/510/test_output/newpynfs.log
> shows that MKFILE fails due to:
> "operation OP_SETCLIENTID_CONFIRM should return NFS4_OK, instead got
> NFS4ERR_CLID_INUSE"
> 
> (Several of the open tests fail for the same reason.)
> 
> This means MKFILE only sent a SETCLIENTID, SETCLIENTID_CONFIRM, then got
> _CLID_INUSE error.  This is what needs to be investigated.
> 
> One thing to try is to use the --newverf option, which forces the use of a
> new verifier every time SETCLIENTID is called.  Though if this works, it
> is just hiding the change in behavior.

Hmm, looks like it's seeing the same issues with NFS4ERR_ACCESS, even with
this option being used:

    http://crucible.osdl.org/runs/848/test_output/newpynfs.log
    http://crucible.osdl.org/runs/850/test_output/newpynfs.log

Bryce