[eepro100] arp/rarp network Issue - Anyone know who to contact about it?

Heflin, Roger A. Roger.A.Heflin@usa.conoco.com
Fri, 1 Jun 2001 10:15:20 -0500


Hello,

We have a problem, as our cluster has gotten larger we have started getting
random 
"No Route to Host" messages.   I am not sure of exactly what the cause is.
It is happening
while a machine is up the entire time, and while all networking is correctly
connected and
nothing is going on (as far as we know).    Trying a few seconds to minutes
later everything
will work (with no system changes).   We have >500 machines and it appears
to have started
being a issue around the time we passed >350 machines or so.    It has
happened to different
machines at different times.  The machines are all using eepro cards
connected to cisco 2948
switches with those connected to a cisco 4006 switch.

I am pretty sure the exact conditions to get the behaviour are: The remote
machine is not in
the local machines arp cache, and the arp request fails for an unknown
reason (packets lost
going out of local machine, remote machine did not answer, remote machine
answered but the
packet did not get back, local machine sent out packet but packet was lost
someway).   So
long as the remote machine is actively used by the local machine we don't
seem to have the
problem, it only appears on new connections.

I really don't think it is a eepro driver problem, it would appear more
likely to me to be a issue
with how the arp/rarp stuff works, but I have no idea who to contact about
this.


					Roger