[c-nsp] About bgp fast-external-fallover

[Oliver Boehmer (oboehmer)]

Hi,

I don't think it's useful for iBGP, assuming you have a redundant IGP path to your iBGP neighbour. There are situations when the route to the iBGP neighbour is withdrawn following the failures, but re-installed a few seconds later after IGP convergence. So if you tear down the session right away, you're loosing traffic for much longer as the iBGP session needs to be re-established.

I'd rather relay on next-hop tracking in iBGP..

	oli

Hiromasa Sekiguchi <mailto:hiromasa.sekiguchi at ctc-g.co.jp> wrote on Monday, March 17, 2008 7:58 AM:

> Hi,
> 
> I configured "neighbor x.x.x.x fall-over" commands.
> I checked that iBGP peer went down immediately after link down on
> iBGP. 
> 
> I think it is useful command on iBGP.
> 
> Regards,
> Hiromasa
> 
> 
> Oliver Boehmer (oboehmer) wrote [2008/03/07 17:19(JST)]:
>> Christian Meutes <> wrote on Friday, March 07, 2008 9:10 AM:
>> 
>>> Hi,
>>> 
>>> --On Freitag, 7. März 2008 12:48 +0900 Hiromasa Sekiguchi
>>> <hiromasa.sekiguchi at ctc-g.co.jp> wrote:
>>> 
>>>> The cisco products have "bgp fast-external-fallover" function.
>>>> It is available on only eBGP, isn't it?
>>>> 
>>>> We'd like to do same behabior like it on iBGP.
>>>> So, is there any solutions?
>>> Many new releases have something similar for iBGP learned prefixes
>>> which is enabled by default. Check for 'BGP Support for Next-Hop
>>> Address Tracking'.
>> 
>> Right. And there is actually a feature "Fast Session Deactivation"
>> which builds on address tracking which takes down the session when
>> the next-hop goes away. Hoever, this is usually a bad idea for iBGP,
>> it's better letting "regular" next-hop tracking invalidate the paths
>> rather than taking down the whole session.    
>> 
>> 	oli