[f-nsp] Tough arp(?) problem

[Gabriel Cain]

Hello all,

So I've got a strange event going on (and I've got a ticket open with
Foundry already about this; nothing useful yet):

I've got a GT EGC16, with some servers connected to it, including a pair of
DNS servers.  It's connected by a dot1q trunk to a Cat6500.

>From time to time, I get this in the logs:
Jul 10 14:35:01:N:L4 server X.X.X.23 dns2 is down
Jul 10 14:35:01:N:L4 server X.X.X.23 dns2 port 53 is down due to server-down
Jul 10 14:37:05:N:L4 server X.X.X.23 dns2 is up
Jul 10 14:37:06:N:L4 server X.X.X.23 dns2 port 53 is up

The servers are in fact not down when this happens.

It seems to go like this:
1.  Server is happy, arp says directly connected, e4/6
2.  down event message shows up,
    "show arp" reveals that the device is believed to be on e3/15 (trunk
    port)
3.  Time passes (a few seconds to a few minutes)
4.  Server returns to up status, arp says it's on e4/6 again.

What I've been able to see is that the arp entry drops off the arp table in
the SI, and shows up again in the arp table as existing on the trunk port.
After a short interval (a few seconds to several minutes), the device shows
up correctly in the arp table as directly connected.

Has anyone seen behavior like this before?  Any ideas?

Gabriel


-- 
Gabriel Cain					Senior Systems Administrator
PopCap Games						  gabriel at popcap.com