[c-nsp] interrupt cpu // processor routed packets

[bill fumerola]

On Thu, Jun 05, 2008 at 10:32:30AM -0400, Rodney Dunn wrote:
> #1 issue with tunnels is usuall a fragmentation reassembley problem.

(damn, i'm usually smarter than this.. :->)

> Watch 'sh ip traffic' outputs for large jumps.
> 
> Clear the counters and capture snapshots of 'sh ip traffic'.

we were already tracking the IP-MIB in cacti, so viewing the ipFrag* and
ipReasm* values made this obvious. good call.

> Also, do sh buff input-interface <name> packet' to see what
> packets are being punted.
> You have to do it against the subinterface if it's a trunk.

sure enough, a bunch of packets a few bytes over the MTU and a few bytes
over the minimum. it would be nice to know through which tunnel they
were coming from. is there anyway to use the memory values from 'sh buff
input-interface' there to display the actual packets (or buffer)?

thanks for your help, this was driving me mad. 

-- bill