4500X problems with high CPU

[spapageorgiou]

Hi all,

We are running into problems with a 4500X. The switch has high CPU and if we add more services to it, it collapses. Let me clarify that:

- We have 189 VLANs and 185 RPVST instances

- About 120 SVIs (VLAN interfaces) with ipv6 *only*. Typical config of the interface follows:

interface Vlanxxx
description servers
no ip address
ipv6 address 2001:xxxx:xxxx:xx::yy/64
ipv6 enable
ipv6 nd other-config-flag
ipv6 nd ra interval 15
no ipv6 redirects
ipv6 mld explicit-tracking
ipv6 mld access-group mld_filter
ipv6 verify unicast source reachable-via rx allow-default
ipv6 dhcp server v6_dhcp
ipv6 traffic-filter SERVERS-OUTV6 out
ipv6 ospf 3323 area 0
counter ipv6

- OSPF, OSPFv3, iBGP.

- ipv6 unicast and multicast

- IP cef contains about 1200pfxs and ipv6 cef contains about 2000 pfxs.

- CPU varies from 30% to 45% depending on the traffic passing by (I think it should be uncorrelated, but CPU correlates strongly with traffic).

When we move our primary ipv6 BGP peering to the 4500X, the load of the CPU goes high. If we add ipv4 on the SVIs the load also goes up and it collapses (100% CPU).

We have about 1.5G of traffic. When we did a local SPAN on one port (mirroring)with 1.5G traffic, the monitor port transmitted only 350Mbps of traffic.... The rest was dropped. Weird and disappointing behaviour for a rather simple HW feature.

Do you have an idea what is the feature that raises up the CPU and makes it correlate with traffic?

Thank you,

Spiros

PS: we are running the latest stable IOSXE

[Iulian Vaideanu]

Maybe this can help you (especially "show platform cpu packet statistics all" and "debug platform packet all receive buffer"):

http://www.cisco.com/c/en/us/support/docs/switches/catalyst-4000-series-switches/65591-cat4500-high-cpu.html

[a.auvinet]

Hi,

Have you still this issue ?

Cdlt,

[spapageorgiou]

Unfortunately yes. We will open a case to Cisco and I hope they will come to conclusion.

Sp

[a.auvinet]

I have also a issue high cpu on my 4500-x If you resolve your problem, can you post this conclusion ? Thanks a lot

[spapageorgiou]

I opened a case to TAC and the conclusion was that there is not particular problem with the setup. The CPU is supposed to be high for our setup. 

We captured packets that go to CPU (there is a mechanism for that in the 4500X) and we examined the packets. Nothing suspicious was found. It is just that we have about 6-7K active users and many vlans and the 4500X just can't handle it.

Sp

[Julio E. Moisa]

Hi

Try to identify the process which is generating the high CPU utilization with the command:

show processes cpu sorted

Also try to open a ticket with Cisco TAC and enable Webex with them to check your device.