Hi Friends,
I've been looking into an issue related to high CPU utilization on a Cisco 2960 stack of switches configured as an access layer switch in an access-distribution block. The initial investigations revealed that there was a rise in multicast traffic on the uplinks and further investigations revealed that certain devices connected on the switch start flooding multicast traffic ( the reason is still under investigation). There is no multicast configured for that Vlan and hence no mrouter is discovered as a result the switch considers the multicast as unknown destination and floods the traffic to all ports on that Vlan. The Distribution switch (WS-C3750X-24S) also receives the flooded multicast traffic and floods it to all its ports in that Vlan.
The problem with the access switch is that these packets are handled by the CPU , see the stats below. CPU spends 44% of its capacity in interrupts which it must ideally not do. The same traffic is handled by the 3750 distribution switch in hardware and no interrupt utilization in the CPU is seen.
%SYS-1-CPURISINGTHRESHOLD: Threshold: Total CPU Utilization(Total/Intr): 91%/44%, Top 3 processes(Pid/Util): 157/36%, 53/2%, 126/1%
As per my understanding the switches handle unknown unicast/multicast in hardware and I suspect something's not right with the 2960, I need some expert info to get to the root of this.
Access switch
Switch Ports Model SW Version SW Image
------ ----- ----- ---------- ----------
* 1 52 WS-C2960S-F48LPS-L 15.0(2)SE6 C2960S-UNIVERSALK9-M
2 52 WS-C2960S-F48LPS-L 15.0(2)SE6 C2960S-UNIVERSALK9-M
Distribution Switch:
Switch Ports Model SW Version SW Image
------ ----- ----- ---------- ----------
* 1 30 WS-C3750X-24S 12.2(55)SE5 C3750E-UNIVERSALK9-M
Thanks in Advance
Umesh
