hi,

i had this happen on an u1 and an u10; my u2 2400 runs solaris 'cos of 
the MP issues in sparc64, but that's another topic.

it was fairly easy to force the system to 'crash' (fall asleep, 
whatever) though i won't call it a 'reproducable' behaviour; in my 
cases it happened when the system was under heavy load (e.g., active 
php/perl running due to dynamic web pages) *and* heavy use of both 
network and scsi i/o -- it's important to mention that the system did 
not hang when only disc *or* net i/o was heavy...

might that be a hint?

rgds,

timo

> On Jun 3, 2004, at 6:05 PM, Martin Husemann wrote:
>>>   Bummer.  Is there anything that can be done to avoid tickling the 
>>> bug?
>>
>> If I'd knew how to trigger the bug (more or less fast) I could 
>> seriously work
>> on fixing it. The problem is not understood good enough and happens 
>> too seldom
>> to give any real advice on how to avoid it.
>
>   Crap. :-(
>
>   Well I'm not sure how I can help.  Has it ever been reported to 
> happen on Ultra1/e systems?  I have a spare one here; I can load it up 
> and use it as a test system to see about reproducing the problem.
>
>        -Dave