[nsp] RIP Update - excessive CPU

[Hroi Sigurdsson]

Gert Doering wrote:

> Guessing: this could be due to excessive numbers of BGP updates coming from 
> WilTel.  A possible reason could be a setup problem in there that causes
> route ping-pong between alternate paths that gets reflected in ever-changing
> eBGP metrics.
> 
> Diagnosing this is hairy, though.  One would need to compare the number
> of incoming BGP messages to "other" upstream ISPs, and then debug into
> more detail to figure out what's happening.

I'd try to RRD/MRTG the below OIDs below and compare the rates between 
upstreams and even iBGP peers. Then check if it correlates with CPU 
usage. BGP dampening may help here in expense of memory usage.
Update rate should be a few per second at most. I'm seeing just under 1 
update per second on three different upstreams.

* bgpPeerInTotalMessages
     "The total number of messages received from the remote peer on this 
connection. This object should be initialized to zero when the 
connection is established."
* bgpPeerInUpdates
     "The number of BGP UPDATE messages received on this connection. 
This object should be initialized to zero (0) when the connection is 
established."

Also take a look at 
http://www.ripe.net/ripe/meetings/archive/ripe-42/presentations/ripe42-eof-bgp/sld145.html 
which talks about troubleshooting route oscillation.

-- 
Hroi Sigurdsson · Chief Knob Twiddler · NGDC A/S
St. Kongensgade 40H · DK-1264 Copenhagen K, Denmark
Phone: +45 3370 1544 · Fax: +45 7025 2687