[j-nsp] bgp peer flapping

[adamv0025 at netconsultings.com]

> -----Original Message-----
> From: Saku Ytti [mailto:saku at ytti.fi]
> Sent: Thursday, April 27, 2017 4:54 PM
> To: Alexandre Snarskii
> Cc: adamv0025 at netconsultings.com; james list; Juniper List
> Subject: Re: [j-nsp] bgp peer flapping
> 
> On 27 April 2017 at 03:59, Alexandre Snarskii <snar at snar.spb.ru> wrote:
> 
> > Having each peer in its own group is one of the best ways to shoot
> > yourself in the foot: instead of all peers with same egress policies
> > share RIB-Out, in this model each one will have its own RIB so memory
> > use will go through the roof.. (well, at least in SP scenario, where
> > most peers need to keep full-view in RIB).
> 
> This should not be true anymore. When looking into ORR I raised this concern
> to JNPR, because in ORR you essentially need group per neighbour to have
> unique view per neighbour, and I was worried about RIB scale exploding.
> However JNPR said that they do not anymore create unique copy of RIB just
> because there is new group. I didn't manage to figure out which release it
> has come and how it works, but supposedly it's before ORR, and ORR came in
> 15.1F4 so at least that and newer releases should be memory-friendly to
> large amount of groups.
> 
Hmm, good point, 
ORR would be impossible with the old behaviour as every time the egress policy changes due to ORR the BGP session from RR to a given PE would be reset. 

adam

netconsultings.com
::carrier-class solutions for the telecommunications industry::