I have 4 separate HP ProLiant DL380 G3 servers with internal Broadcom
gigabit NICs. Up until recently they were all NW65 SP5. They were all
stable until late summer when we upgraded to ZFD7 SP1, and replaced
McAfee ViruShield with Sophos AV.

Gradually they all started doing the same thing. At random periods each
one of them will start responding more and more sluggishly until it's
totally unresponsive. At the server console the server appears to be
responsive, it's just incommunicado. However, attempting to unload any
module will basically freeze the server. The only solution is to
power-cycle the box and reboot.

Eventually I was able to watch this process live and in color. The only
obvious symptom I've been able to find is a steady increase in the
"Receive discarded, no available buffers" statistic (Monitor, LAN driver).

Here's what I've done so far: Updated all HP firmware, both system BIOS
and other hardware like NICs, SCSI drives, controllers, etc. All now
fully up to date. Updated all HP drivers to most recent. Upgraded to
NW65 SP6, with all recommended post-SP6 hotfixes as of about 10/1.
Finally, based on various Novell support documents I've doubled the
default value for the Minimum Packet Receive Buffers. That last one
seems to have helped a little bit.

HP has no ideas other than the stuff I've already done.

A friend nearby is using Dell servers with Broadcom NICs, and she tells
me she experienced almost exactly the same series of problems last year.
She did the BIOS/firmware and driver updates, but *not* SP6 - her
problem went away.

Anyway, the slowdown/freeze problem continues, I'm pretty much at wits
end, and the boss is starting to look at me funny. Any ideas or



P.S. Several other HP Proliant servers with slight variations (slightly
different model, slightly different Broadcom NIC, etc.) are not
experiencing the problem! All have the same BIOS/firmware updates,
driver updates, and NetWare SP level (NW65 SP6).