[evla-sw-discuss] ACU hanging issue

[James Robnett]

I've been looking at traffic and I'm relative convinced that
this was not caused by the Operator screen but by the network
switches themselves.

IGMP is the protocol used by computers to register to receive
multicast traffic.

CGMP is a Cisco protocol used by switches (and routers) to share
multicast registration and multicast route (mroute) information
(the rest of the world calls this IGMP snooping).

Normally the background CGMP traffic across all the EVLA VLANs
(around 50 vlans) and switches (around 100 switches) is typically
around ones of packets per second.

Dumping all the traffic to antenna-2 over the past hour shows
250K packets of which around 200K were CGMP.  So around 55 packets
per second of CGMP traffic.  At least 10x what I'd expect.

The peak rate during that period was 163 packets/second, min rate
was 1.  It's pretty variable.

My guess is we had a burst above 400 or so packets per second
that knocked the ACUs over.  It will probably happen again but
could be hours, days or weeks before we see it.

I'll figure out what's going on.  I'm suspicious the bloody IOS
upgrade Cisco foisted on us is involved at some level.

james