[j-nsp] Resolving BGP routes using LDP tunnels

[Abhishek Verma]

Thanks Piotr!

The reason i wasnt sure about BGP using the LDP labels was because LDP
LSPs are all best effort LSPs and are not really traffic engineered.
So i wasnt sure if BGP is going to use them.

But it seems that BGP will pick up anything thats placed in inet.3

Thanks for the clarification!


On 7/13/06, Piotr Marecki <peter at mareccy.org> wrote:
> On Thu, Jul 13, 2006 at 05:18:38PM +0530, Abhishek Verma wrote:
> > Hi,
> >
> > I have a doubt with using BGP and it recursively resolving the
> > next-hops using MPLS tunnels.
> >
> > R1 -- R2 -- R3
> >
> > I am running LDP between R1, R2 and R3 and R1 is aware of R3's
> > loopback address to mpls label binding because of LDP.
> >
> > Now, i am running BGP between R1 and R3. There are some routes
> > advertised by R3 to R1, with R1's NEXT_HOP.
> >
> > Is it possible for R1 to use the labelled path from R1 to R3 when
> > recursively resolving the next-hops for the BGP learnt routes.
> >
> > I know that this happens when there exists an RSVP tunnel from R1 to
> > R3. My question is that will this also work for LDP exchanged labels?
> >
> > Thanks,
> > Abhishek.
>
> In general answer is YES. By default , RSVP IP address ( "to" statement ) and
> LDP LSP's towards primary /32 on loopback will be installed into router inet.3
> table . During recursive lookup for BGP NEXT_HOP , router from inet.0 and inet.3 are used ,
> and if there is preference/metric draw , inet.3 routes are preffered ( which is rather unusul ,
> given that RSVP has 7 and LDP 9 ). So , assuming that your configuration is pretty straightforward ,
>  you will be able to use LDP LSP's out ouf the box :> ( though , if RSVP will be active in the same time ,
> by default RSVP LSP will be used because of protocol preference )
>
> Piotr Marecki
>
>