[f-nsp] Tough arp(?) problem

[Gabriel Cain]

Niels Bakker wrote:
> 
> Can you elaborate on your network topology?  Is it possible that you 
> have a loop behind 3/15 that would cause a station movement for the 
> server?

It's a very straight-forward topology:

Cat6500
  |
SI-GT
  |  \
DNS1  DNS2

3/15 is part of a four-port dot1q trunk (3/15-16, 4/15-16).   Each of the
two DNS boxen are directly connected.  dns1 = e3/6, dns2 = e4/6

Further info, when this occurs (as it just did a few minutes ago), the
catalyst still thinks that the foundry knows where it is.  Example:

During Outage:
Catalyst:  X.X.X.22 (00:30:48:57:1b:6a) is on Vlan10
ServerIr:  X.X.X.22 (00:30:48:57:1b:6a) is on GigabitEthernet3/16

Now (no outage):
Catalyst:  X.X.X.22 (00:30:48:57:1b:6a) is on Vlan10
ServerIr:  X.X.X.22 (00:30:48:57:1b:6a) is on GigabitEthernet3/6

Additionally something amazing happened, when I pinged the servers from a
box in the same subnet, they came back up shortly after the ping.  As though
the SI were suddenly made to go arping and say, "Oh, there they are."

Thanks!

Gabriel