High Proc CPU on Line Card

[Sgt Rock]

Have a interesting problem on a 12008 (old I know so no TAC help). On one of our Single Port Gigabit Ethernet/IEEE 802.3z cards I am getting VERY high CPU usage with no ascertainable reason. The "execute-on slot 2 sh proc cpu " shows 100%/99%/99% across with the actual PID's taking 0.4% total. So I thought it might be the memory, the largest block is still fine, then I thought of CEF causing the issue really didn't look bad at all.

Any other ideas out there short of dropping some $$$ and getting a new chassis with some fabric enabled sups and blades?

Thanks a ton!

 

[Sgt Rock]

OK as I have looked at it it seems it mite be related to incoming packets. I can send out easily 300k packets a sec (note not Bandwidth but packets/sec) but it seems if it starts approaching 180k inbound I start hitting the Proc wall. The routing/switching on the card should be minimal on the inbound packets as CEF and specific route Maps I have setup should already have made those decisions. It could be the application of said maps to that interface though that is causing the Proc usage in determining where to send the packets. But then I would be seeing that on all my Line Cards.


Any further ideas would be welcome, and encouraged.

Thanks

 

[Impulse25]

Which routing engine are you using in that chassis? Sounds like you are just hitting the ceiling in terms of raw packet throughput if you have an older route engine.

 

[Sgt Rock]

We are using the GRP-B on not the PRP's but if the Route Process Module is limiting then why are my other 5 line cards not hitting the same wall? I do see where that may be an issue, so I have put in to buy some 6509's with the 720 SUP to get some fabric enabled cards working as I do think that this piece of Iron is on its last legs.