[c-nsp] 2821 BGP CPU Usage

Brian Desmond brian at briandesmond.com
Mon Aug 28 22:22:27 EDT 2006


I have a pair of 2821s w/ 768MB in them and a 4 port FE HWIC. 

 

#1 has a session to ISP 1 with about 75K routes

#2 has a session to ISP 1 with about 75K routes

#2 has a session to ISP 2 with about 192K routes (full table)

 

They've been humming along fine for almost a year like this, they're
moving virtually no traffic (<1mbps). No features running really, very
short ACLs on the interfaces plus prefix-lists in and out on the bgp
peerings.

 

#1 is running 12.4(9)T

#2 is running 12.3(8r)T7

 

I setup an iBGP peering between the two routers this evening, and after
perhaps half an hour the CPU on #2 started spiking to 99% every minute,
the offending processes were BGP Router and IP RIB Update. #1 started
doing the same thing, so the iBGP peering went away. 

 

The only bug I found that's remotely close is CSCse30673.

 

Any thoughts?

 

Thanks,

Brian Desmond

brian at briandesmond.com

 

c - 312.731.3132

 



More information about the cisco-nsp mailing list