[c-nsp] Huge SP CPU usage spikes 100%

[James Bensley]

On 1 March 2018 at 09:53, james list <jameslist72 at gmail.com> wrote:
> xxx#show ibc
> Interface information:
>         5 minute rx rate 944000 bits/sec, 793 packets/sec
>         5 minute tx rate 25000 bits/sec, 37 packets/sec
...
>         2467023087 Packets out of 554699386 CEF Switched, 0 Packets out of 0
> Tag CEF Switched
>         3916625157 Packets Fast Switched
...
>         Potential/Actual paks copied to process level 228808364/225833216
> (2975148 dropped, 265 spd drops)
...
> MISTRAL ERROR COUNTERS
...
>         2974883 total packets dropped on throttled interfaces (2954630 low,
> 16704 medium, 3549 high)


>> On 1 March 2018 at 08:29, james list <jameslist72 at gmail.com> wrote:
>> > Dear experts,
>> > has anybody experienced a 100% SP CPU usage on C6500-Sup720
>> > (12.2(33)SXI5)
>> > with a lot of interrupts ?
>> > The main process is Heartbeat.
>> >
>> > Cisco TAC is struggling in having an idea to sorting out the issue, they
>> > are working since 3 days on it..
>> >
>> > STP is stable, no mac moving, no real issue found… maybe somebody
>> > experienced the same due to something in particular?

I've compared to a similar box I have, it has less control-plane
traffic than yours it would seem. You have a decent amount of dropped
packets which I guess to be expected if you have sustained 100% SP CPU
utilisation.

Do you have a lot of spanning-tree instances, HSRP/VRRP, multicast
(various other control-plane stuff) running on this box?

Have you confirmed there are no layer 2 loops?

If you run "show mls statistics" several times do you see any error
stats climbing very quickly?

I would also run a netdr capture and check if the traffic in their is
normal for your network.

Cheers,
James.