Catalyst 4500 High CPU caused by Call Manager MOH multicast traffic

[craigallen]

I have found an as yet unregistered. Does anyone run 4500's with Call Manager using MOH in Multicast operation? Do you experince high CPU, but show platform traffic stats shows only 1 - 5 % traffic loading?

[Prashanth Krishnappa]

Can you provide more details? Have you configured IP multicast routing on the switch? Post output of sh mod, sh run. When the CPU is high, capture an output of sh platform health.

[craigallen]

This is on TAC case 601584863. I configured unicast music on hold on the Call Managers and immediately the CPU load ropped to less than 10%. The packet rate from the CPU SPAN session was only 178 frames per second UDP16384 to multicast address 239.1.1.1 and it was causing the CPU to run at 60 - 70% util 24 hours a day. During peak hours the average would rise to 75% the show platform traffic stats would show chassis traffic at 1% and approxiamtely 2500 frames per seoncd.

[Prashanth Krishnappa]

I see "bad hop count" increment in the output of "show ip traffic". This indicates the TTL is currently set to 0. You could capture a sniffer trace of the CPU port to verify if this is the case.

http://www.cisco.com/warp/public/473/82.html#understand

Set a TTL setting on your MOH server to a higher value say 5 or 6.

[bbaillie]

Muuticast Packets with TTL = 1 are handled by the CPU because they require special handling. Normally these packets are switched by the ASICS therfore not impacting the CPU utilization. Setting the TTL to a higher value is your best plan of attack in this situation.