ESIX "pauses" and panics

[Dave Shepperd]

We have 2 systems running ESIX 3.2. One is on a Micronics 486/25 and the other
on a Club 486/33. They both exhibit a problem where once every few minutes,
outbound network traffic will pause for a few seconds. The problem is worse
on the 486/33 being that the pauses are more frequent and last longer. When
we were running ESIX on a 386/25, noone seemed to notice any such problem
(that's not to say it didn't do it, only that no one remembers that happening).

In an effort to try to isloate what might be causing this, I created a 2 node
network consisting of the 486/33 and an X-terminal (the two nodes are connected
together with a single 10 foot piece of cheapernet cable and terminated at each
end). I logged into the ESIX system from the X-terminal via telnet and ran the
following program:

#include <stdio.h>
#include <sys/types.h>
#include <time.h>

main()
{
   time_t t;
   while(1) {
      t = time(0);
      fprintf(stderr,"The time is %s",ctime(&t));
   }
}

and watched for the pauses. I anticipated that I could zap processes
while watching for the pauses until perhaps things cleared up. What
happens instead, is the kernel will panic after a few minutes OR it'll
just lockup (allows no network activity at all; will echo characters
typed on the console, but will not act on them).

The question is, has anyone else seen this behavior and, better still,
is there a fix?
-- 
Dave Shepperd.	    shepperd at dms.UUCP or motcsd!dms!shepperd
Atari Games Corporation, 675 Sycamore Drive, Milpitas CA 95035.
Nobody knows what I'm saying. I don't even know what I'm saying.