On Fri, 21 Aug 1998, Jonathan Stone wrote:

> I beleive there were TCP bugs in 1.3_BETA which could be causing, or
> contributing, to the poor TCP throughput.  These got better in 1.3.2.
> I would recommend upgrading both machines to 1.3.2 if at all feasible.

Yep. Will do that as soon as possible. Haven't got physical access to the
pmax at the moment though. And the battery in the pmax is broken so I
can't reboot it remotely.

> If nothing else, you can install a 1.3.2 kernel and see if that helps.
> I expect it would help significantly.

I've just upgraded the hp300 to 1.3.2. It didn't do any difference. Still
getting lots of "le0: lost carrier" messages... :(

The hp has an AUI with a transceiver connected to it. And as soon as I put
some load on the segment it prints the error messages. And the error count
increases if I do a "netstat -I le0 -w 1", about 1 every 2 or 3 seconds as
long as there's heavy traffic on the segment.

And the pmax outputs these messages maybe once every ten seconds, 24 hours
a day. But there's a *lot* more traffic overall on this network segment so
that's maybe understandable. But it's just as annoying...

> >And if I try to ping the machine when there's heavy load on the net,
> >it looses a lot of packets. The i386 machine on the same segment works ok
> >regardless of the network traffic.
> 
> To me, that sounds like a completely different problem.

Maybe so. But what's funny is that both the hp and the pmax is showing
exactly similar behaviour when I'm stressing the ethernet a bit. 

> PS: precisly what version of NetBSD (or OS) is the i386 running?

NetBSD 1.3.

--
Jorgen Pehrson                  HP 9000/380 (NetBSD/hp300 1.3.2)
jp@spektr.ludvika.se            DECstation 5000/200 (NetBSD/pmax 1.3)
http://spektr.ludvika.se/museum PDP11/83 (2.11BSD) VAX2000 (NetBSD/vax)