[c-nsp] Good practices for peering

[Stephen J. Wilcox]

On Mon, 2 Jan 2006, Danny McPherson wrote:

> On Jan 2, 2006, at 5:52 PM, Richard J. Sears wrote:
> 
> > Thanks Arnold -
> >
> > I assume that if I already have those routes in my IGP via some  
> > other way
> > that I can safely disregard the set-next-hop-self...?
> 
> See my earlier response for some additional insight (our messages
> likely passed on the wire :-).
> 
> However, in short recall that the motivations are quite different when
> applying this to iBGP peers (which Arnold mentions) versus eBGP
> peers, both of which I recommend.   A quick summary:
> 
> EBGP next-hop-self configuration:
> 
>   o static setting
>   o ensure NBMA connectivity

there have also been issues in the past when more specific prefixes of the
next-hop lan have entered the routing table via bgp causing traffic to head the 
wrong way thro your network

eg IX uses 10.0.0.0/23 ethernet lan. anomolous 10.0.0.0/24 is received in bgp 
from a peer, traffic will head towards that peer not the IX.

this also can be mitigated by filtering IX prefixes in BGP

Steve

> 
> IBGP next-hop-self configuration:
> 
>   o update packing
>   o remove external subnets from IGP (lots of benefits)
>   o TE, etc..
>   o NBMA depending on where next hop is derived
> 
> There are some tweaks and caveats that need to be considered with
> things like route reflection (next-hop-self doesn't apply to reflected
> routes as a default behavior) and confederations, etc..  And I'm pretty
> sure there are other reasons for both of these, just  haven't put a  
> whole
> lot of thought into it at the moment :-)
> 
> -danny
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>