🚀 go-pugleaf

dg0951@cec1.wustl.edu (Daniel Geist) writes:
>Thanks to those that replied (And there were many) to summarize
>the reponses:
> [3 suggestions deleted]

4.) Be sure the sun is set up with a brodcast address set to all bits 1 if
    your ethernet does not contain just suns.

We had severe ARP-storms a few days ago. They were triggered by a sun with
a broadcast adress set to all bits 0 (standard sun setting).  I used
'etherfind' on some other sun (with correct broadcast setting) to analyse
the disaster: lots of ARP-backets, lots of other broadcast traffic (more
than 1000 times as much as usual [maybe broadcast resending ? Some
machines sent out 10 to 50 broadcast RIP and rwho pakets in a second.
Usually the rates are one packet in 30 seconds resp. one packet in 3
minutes]). All the problems went away a soon as I fixed the broadcast
address of the offending sun.  We have a mixed envronment: Sun3, Sun4,
Apollo Domain, PCS Cadmus, IBM-AT, MacIntosh and a VM/XA system.  The
'storms' does not show up here when we used class C networks without
subnets. It just started since we use a registered subnetted class B
network. All Suns run SunOS 4.0.3. The other machines use the BSD 4.3
TCP/IP code and other stuff.

/ UUCP:    lehners@uniol              | Joerg Lehners                  \
|       ...!uunet!unido!uniol!lehners | Fachbereich 10 Informatik ARBI |
| BITNET:  066065 AT DOLUNI1          | Universitaet Oldenburg         |
\ Inhouse: aragorn!joerg              | D-2900 Oldenburg               /