We have a OES (SP2, fully patched) server with NSS volumes. Every few
days (sometimes even a couple of weeks) NCP stops responding. Users
can't browse directories, open files, login or do pretty much anything
until someone goes onto the server and does /etc/init.d/ndsd restart .
Even getting a ssh onto the server is very slow, and the console itself
is sluggish - can't even do a "top" to see what's going on.

There's nothing obvious in any of the logs (/var/log/messages,
/var/log/novell/ncp2nss.log etc). However, the ncpserv.log is full of
things like this:

[! 2007-04-19 03:46:47] ncp_idbroker_handler: (verb=0 len=16 ver=1)
handler failed rc = -765
[! 2007-04-19 03:46:47] ncp_idbroker_handler: (verb=0 len=16 ver=1)
handler failed rc = -765

(this repeats many times, although it seems to happen in "chunks")

Also:

[! 2007-04-19 07:53:13] NCPMapGUIDToUID: UID for this object (850e) is
0, returning -601
[! 2007-04-19 07:53:13] ncp_idbroker_handler: (verb=5 len=16 ver=1)
handler failed rc = -601
[! 2007-04-19 07:53:13] NCPMapGUIDToUID: UID for this object (850e) is
0, returning -601
[! 2007-04-19 07:53:13] ncp_idbroker_handler: (verb=5 len=16 ver=1)
handler failed rc = -601

None of these errors seem to relate to the time when NCP hangs.

We have another OES server in the same tree with _no_ NSS volumes, and
it runs flawlessly. Interestingly, I had exactly the same problem with
that server in a "previous life" - it originally had NSS volumes, we
couldn't make it stable, so we flattened it and rebuilt it without NSS.
That was over a year ago, and I was hoping all the problems with NCP/NSS
had been solved by now.

Can anyone suggest what we can do to stabilise NCP/NSS?