I'm evaluating NW65SP1b for use as a NFS server and ran into two problems so far. This posting is dedicated to my "rm -rf" problem.

When a large tree of files and folders should be deleted using "rm -rf", sometimes the search for files seems to be aborted too early, leaving files in the folder to be deleted. The deletion of such a folder, of course, fails. The manual states that this is a known issue for NFS3 over TCP. However, I'm experiencing this problem in all kinds of protocol combinations I tried. To reproduce:

$ tar xvfz bash-2.05b.tar.gz; rm -rf bash-2.05b
rm: cannot remove directory `bash-2.05b/tests': Directory not empty

In this situation the "tests" folder retains four regular files. Rerunning the same "rm -rf" again and again deletes more and more files and eventually succeeds. This is not an acceptable workaround as the tree removal is usually called from within a script not under my control. It is not a permission problem. It does not happen on a local drive. A tcpdump shows the error is returned from the server's RPC so I do not expect it is a NFS client cache issue.

How can I make "rm -rf" work?

the client is a
$ cat /etc/SuSE-release
SuSE Linux 9.0 (i586)

and the server is a
nw65ossp1b.exe NetWare 6.5 OS SP1.1 Overlay CD 18MAR2004
nw65prodsp1b.exe NetWare 6.5 Products SP1.1 Overlay CD 18MAR2004

the following mount options did not make a difference
nfs (rw,nfsvers=2,actimeo=0,tcp,noac,noacl)
nfs (rw,nfsvers=2,dirsync,tcp,sync,actimeo=0,noac,noac l)
nfs (rw,nfsvers=2,dirsync,udp,sync,actimeo=0,noac,noac l)
nfs (rw,nfsvers=3,dirsync,udp)
nfs (rw,nfsvers=3,dirsync,udp,rsize=512,wsize=512)
nfs (rw,nfsvers=3,dirsync,udp,rsize=8192,wsize=8192)