[c-nsp] HSRP Flapping Due to CPU Spikes

[hjan]

alaerte.vidali at nsn.com ha scritto:
> Hi Gianluca,
> Thanks for your information. (sorry the delay between one message and
> other, out of office)

Sorry foir the delay but i'm out of office due to honeymoon vacation :)

> I think this may be a central point of the instabilities we are
> observing on 7600. I am trying to understand why ""mls rate-limit
> unicast ip icmp unreachable..." help in your case.

This could be long...however i try to summarize below.
The traffic punted to CPU has been identified as user traffic. This
traffic is normally forwarded through the router in hardware and not
transferred to the RP. During transient conditions it is possible to see
traffic locally switched on the same port. Local switching on the same
port causes the router to send these packets to the RP for processing
and to originate ICMP redirect messages (if necessary).
The command "mls rate-limit unicast ip icmp redirect" allows to
rate-limit the packets sent to the RP.

> 
> Please if you have time let me know:
> 
> -Did you use commands to verify mistral asic drops?

Yes.
show ibc | in LBIC at some point i can see:
sho ibc | in LBIC
   LBIC RXQ Drop pkt count = 65535            LBIC drop pkt count  = 0
   LBIC Drop pkt stick     = 0
so drops occour.


> -Is the traffic punted to RP data traffic? 

Yes.

> -Are you using logging on Access Lists?

No.