EIGRP Neighbour Drop High CPU

[Keith Alman]

Would anyone know if it possible to prioritize a certain process, or for that matter only allow certain processes to consume X amount of CPU?

Consider the following scenario. Two 7304’s with NSE100’s running full BGP tables with iBGP between them. iBGP peering is established using loopbacks. The loopback addresses are passed via the physical infrastructure using an IGP (EIGRP).

At the moment, when one of the eBGP peers flap, the BGP scanner process hits the CPU hard, staying at 100% for over 60 seconds. During this time EIGRP cannot process the hello packets and looses neighbours, which wreaks havoc – as one could imagine. How could I get over this? Would using the scheduler command set to lower intervals help?

Thanks in advance,

Jansen

[Keith Alman]

The scheduler doesn't really seem to help that much

[vladrac-ccna]

this might help you:

"Troubleshooting High CPU Caused by the BGP Scanner or BGP Router Process"

http://www.cisco.com/warp/public/459/highcpu-bgp.html

HTH,

Vlad

[Keith Alman]

Thanks for the link. I've already implemented all relevent suggestions. It does make BGP propagation faster, lowring the time BGP hogs CPU, albeit not enough.