[c-nsp] IOS XR BFD

[Ivan Pepelnjak]

I've been planning to document the shortcomings of "Fast Peering Session
Deactivation" for a long time; thanks for the nudge.

Summary: following an interface loss (on the BGP router) in an OSPF or IS-IS
network, you might lose the route toward your BGP neighbor until SPF is run,
resulting in BGP session loss.

I've written an article in our wiki for those of you who want to know more:

http://wiki.nil.com/Aggressive_BGP_fall-over_behavior

Ivan
 
http://www.ioshints.info/about
http://blog.ioshints.info/

> -----Original Message-----
> From: Mateusz Blaszczyk [mailto:blahu77 at gmail.com] 
> Sent: Tuesday, July 07, 2009 4:31 PM
> To: Ivan Pepelnjak
> Cc: cisco-nsp at puck.nether.net
> Subject: Re: [c-nsp] IOS XR BFD
> 
> Ivan,
> 
> >
> > BTW, even the more "traditional" fast convergence 
> techniques (internal 
> > BGP fast fallover) might be too aggressive and do more harm 
> than good.
> >
> 
> Could you elaborate little more on that?
> I thought it would be a good idea (e.g. neighbor X fall-over
> route-map) to drop BGP session with a neighbour that suddenly 
> "dissapeared" from the network.
> In my scenario I am concerned that the scanner doesn't 
> invalidate the routes because I have catch-all aggregate 
> covering all my NHs floating there (I can't have full table 
> so I have 0/0 from upstreams so I need the aggregate for my 
> routes) so in other words it takes 3 minutes to close the 
> broken session.
> 
> Best Regards,
> 
> -mat
>