help: random crashing of DS5000 running ULTRIX 4.1

[George Robbins]

In article <1606 at usage.csd.unsw.oz.au> mcba at newt.phys.unsw.OZ.AU (Michael C. B. Ashley) writes:
> Hi,
> 
> This is a rather long message describing a problem I have with a
> machine crashing. If anyone could shed some light on a possible
> solution, I would be most grateful.
> 
> I have a DS5000/200PX running ULTRIX 4.1 (Rev. 52), and the machine
> crashes an average of once a day...
...
> No error messages are observed in the output of /etc/uerf. The machine
> (and memory and disks) pass every diagnostic in /usr/field, and I have
> run the hardware (V5.3) ROM tests for days at a time without picking up
> any errors. Last week the system board was replaced, however, the
> problem remains.
> 
> Once I noticed a message similar to "swap error" appearing in the
> Session Manager message area at the instant of a crash...

Why don't you attach a decwriter or some similar device and set up
for hard-copy console mode.  Then you can put in crontab periodic
executions of pstat, vmstat and ps to see how things go downhill
leading to the "crash".

You may find that your swap space is eroding or mountd is cancerous
or something else obvious that you lose from the crt screen.

> The machine will run without crashing if I disconnect the ethernet. The
> crashes aren't related to some user's program, since there aren't any
> users other than root at the moment.

Hmmm...

>   (2) Since the crashes appear to be related to the ethernet, maybe I
>       need the "ln*.o kernel fix" that has been mentioned recently
>       with respect to using tcpdump and LAT with ULTRIX 4.1. Note
>       that our ethernet is teaming with exotic packets from all sorts
>       of machines, and regularly crashes a couple of VT1000's we have
>       in the building (they die with "illegal opcode 28", despite a
>       recent ROM upgrade, but that is another story ...).

Well, you can always try to talk the support center into analyzing a
crash dump...

-- 
George Robbins - now working for,     uucp:   {uunet|pyramid|rutgers}!cbmvax!grr
but no way officially representing:   domain: grr at cbmvax.commodore.com
Commodore, Engineering Department     phone:  215-431-9349 (only by moonlite)