BGP bestpath selection question

Hi guys,

I have a network with BGP and 3 uplinks. The route flap dampening feature isn't configured. Recently there was a situation, when session with the ISP#1 went down (there was a L2 issue and the hold time expired) and most of the traffic started flowing through ISP#2 and ISP#3. However, the mentioned disruption was quite short. When the session came up again, the routes were not recalculated as fast as I would expect. I mean, the the traffic was still going through the other 2 uplinks, instead falling back to the ISP#1.
It took a while to fall over to ISP#1. Can anybody explain?

Find more posts tagged with

Comments

Chris_

How long is 'a while' ? BGP is not exactly the quickest protocol in the world!!

jason_lunde

This is by design for BGP. Imagine if every node on the internet ran eigrp and networks were converging as fast as it could detect the network up or down (this is not even considering that BGP has route dampening which lessens these effects). But every time you had some reckless admin bouncing his connection it would cause this massive convergence wave over the internet backbone. This is why BGP has a fairly lengthy scan time, and is also why there are things like route dampening and why ISP's summarize. These items together help stabilize the global routing table and keep less knowledgeable people who control BGP routers from ruining it for everyone else.