[c-nsp] Sporadic loss of LDP neighbor ...

Mark Tinka mtinka at globaltransit.net
Fri Dec 16 12:04:23 EST 2011

On Friday, December 16, 2011 10:31:21 PM Garry wrote:

> I've seen high CPU before, but never in time to discern
> whether the CPU was cause or effect ... just this
> afternoon I was able to catch one of the outages in time
> to cross-check multiple places, mainly the logs and cpu
> history, which clearly showed that the egg was there
> before the hen - or rather, 100% cpu for ~2min followed
> by the LDP (and other) outages ... problem is, I can't
> yet pin-point the cause of the CPU load - guess I will
> have to set up a cron job to pull "show proc cpu sort
> 5min" outputs every couple minutes and check which
> process is the cause for the cpu load ... hopefully ...
> (even as I prepare to drop the 7200's from the essential
> places, I even see 10% cpu at the same time on the ASR
> routers, which is pretty high compared to the ~1% they
> usually have ... so I want to solve the problem cause,
> not the effect ...)

Yes, the LDP session failure hints toward high CPU state 
moments before the LDP sessions go down. Other (routing) 
protoocols are also likely to be affected.

In our case, we've seen this before for two reasons:

	1. Removal and insertion of compact flash card
	   causes high CPU utilization.

	2. Heavy QoS on the 7200.

As you say, figuring out what's killing your CPU will likely 
be the solution to taming this issue.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20111217/e56eac3c/attachment.sig>

More information about the cisco-nsp mailing list