Foundry sflow idiosyncrasy

For those in the sflow community using Foundry gear:

I have been trying to figure out why some of our removed gear was not
timing out and still showing a red square in the status column (normally it
just disappears after a while). I was about to delete stuff out of
/usr/local/inmon/server/state but wasn't sure if that would produce
unpredictable results so I investigated further.

The answer is that our equipment was redeployed elsewhere with a different
IP address, but the Foundry switches (FES2402 in these cases) hang on to
the old IP address for sflow agent IP address even through an IP address
change and reboot. Traffic server is still receiving samples from the
switches at the old IP address but cannot reach them for PING or SNMP at
that old address.

To see if this is plaguing you, do a show sflow and show ip on your switch:
telnet@6-12-1-1964>show sflow
sFlow services are enabled.
sFlow agent IP address: 10.106.12.1
...
telnet@6-12-1-1964>show ip
      Switch IP address: 10.103.8.1

The fix is toggling sflow which I'll do rather verbosely to show that it
really is the fix:
telnet@6-12-1-1964#conf term
telnet@6-12-1-1964(config)#no sflow enable
telnet@6-12-1-1964(config)#show sflow
sFlow services are disabled.
sFlow agent IP address: 10.106.12.1
...
telnet@6-12-1-1964(config)#sflow enable
telnet@6-12-1-1964(config)#show sflow
sFlow services are enabled.
sFlow agent IP address: 10.103.8.1
...

So if you have foundry gear that won't go away or shows up in the wrong
place, check the sflow agent ip address.

Regards,

Kevin.
Received on Wed Apr 12 10:50:15 2006