[comp.sys.apollo] DN10000 Ethernet Controller

system@aurum.chem.utoronto.ca (System Admin (Mike Peterson)) (06/01/91)

The Hotline says that the PROM upgrades for the DN10000 Ethernet card
are ready, and should correct all the TCP/IP problems (hesitations,
delays, nodes dropping of the net, nodes going comatose) - I will
believe that when I see it !!!
Apparently you have to contact your local support people to get the
PROMs ordered - WHY AREN'T THESE BEING SENT TO ALL SITES AUTOMATICALLY ???

We are also running SR10.3.0.6 on our DN10000, which is supposed to have
a software "patch" around the Ethernet problems - it may be helping
avoid the hangs (though we had one with the patch), but the
hesitations and delays are so bad that you have at best a 50% chance of
successfully doing a telnet/rlogin/ftp/rsh to/from the DN10000, and
sessions are constantly timing out.

P.S. Thanks for the posting about NFS 2.2; I would have already installed it
but I never received it for our DN10000 (someone at Apollo tells me it
is on the current software list though). We have enough TCP troubles without
that. I agree completely with the poster of that article -- "how does
stuff like that get out?" --- and I'll add "who beta-tested it and thought
that it worked"?
-- 
Mike Peterson, System Administrator, U/Toronto Department of Chemistry
E-mail: system@alchemy.chem.utoronto.ca
Tel: (416) 978-7094                  Fax: (416) 978-8775

attfield@bcrka269.bnr.ca (Ed Attfield) (06/04/91)

  We've tried the prom upgrade with various jumper combinations.
Actually, it came as a complete replacement board - courtesy of our 
apollo 10k rep: Dan Murray.

  The one we got seems to be defective somehow. Perhaps the fixes don't
work in real life?

  It fails repeatably on the following test:

    telnet to the 10k
    /etc/lcnode

  -- it types about 5 lines worth of the lcnode, and hangs the telnet session.

  We took it out, and the board has gone back to the lab for more work.

  We are getting good results with the crowbar patch contained in 10.3.0.6
so that the node rarely gets disconnected from the network (unless it does
the VME bus time-out crash ...)

*
* Ed Attfield                                      email: attfield@bnr.ca
* Bell Northern Research                           phone: 613-763-3102