We upgraded to NW 6.5 SP4a (we're also running Veritas Backup Exec 9.2)
and we've had problems with memory leaks ever since. Right now the
server memory statistics show 9% Allocated memory pool and 0% Cache
buffer memory (we're preparing to reboot the server before it crashes).
After a reboot the Cache buffer memory shows around 48% and gradually
shrinks each day.

We had a couple of abends that listed CONNMGR.NLM v5.60 as the cause, but
we normally try to reboot before we get an abend.

We're running on an HP DL380 Proliant (dual processor, 4gb ram) connect
to an MSA1000 SAN device.

I've tried making the changes recommended by error messages (added -u
parameter to the SERVER.exe line in autoexec.bat), which helped but
didn't solve the problem.

Downloaded the seg.nlm from Cool Tools (have the .csv results from that
if someone thinks it would help to analyze it).

Also changed the Cache Memory Threshold (in tsa.cfg) from 10 to 5 and now
to 2. The seg.nlm is recommending using -u705000000, so we'll try that
for this reboot.

Seg currently shows suspect NLM's to be portal.nlm, aprlib.nlm and
tcp.nlm. Watch nlms are: httpstk.nlm, connmgr.nlm and ncpip.nlm.

We've installed the latest Novell drivers from HP. I just installed the
latest NWFTPD patches (we do quite a bit of FTP transfers to our server)
just in case.

I'm stumped.

Wes Harper
Network Administrator
Pioneer Telephone Cooperative, Inc.