[c-nsp] 6500 troubles. What caused these packets to be punted? Is that even what went wrong?

[Jim Devane]

Hello,

    I have a 6509, SUP 2A, PFC2, MSFC2, running 12.1.22 and a few 6408A and 
6348 modules. 

I turned up a customer on a gig port on a 6408. 
The only config on the port was a description and ip addr. 
Running BGP to the customer. 
I am sending a default route and they are advertising 2 routes to me. 

They are sending small UDP packets (60 bytes) at roughly 10K -30 K pps for 
about 8-20 Mbps. 
The moment they began sending traffic the CPU jumped from 11% to 20% and the 
chassis as a whole began to experience packet loss. 

This traffic level, and even the CPU level seems like it should have not 
caused this dramatic of a problem. When the packets were stopped form this 
customer all fell neatly back into place.

Our Cisco cpu normally handles well over this number of pps and hits' a 
daily high of about 25% for CPU. 

Any suggestions (yes, besides UG the code and getting a 720) on why this 
happened? I am not sure why the packets would be punted. I am assuming this 
is why the CPU rose so much. There is no ACL, no QoS that would caused the 
packets to be processed switched as far as I am aware.

Would very much appreciate any steers to resources, suggestions, silver 
bullet solutions.

TIA,
jim