[j-nsp] BGP RR in MPLS VPN

Ihsan Junaidi Ibrahim ihsan at isp.time.net.my
Sat Feb 10 17:16:12 EST 2007

Hi all,

Bringing up an old topic. :)

I'm having problem creating an LSP on one of my RR to it's own lo0. The reason I'm doing this is to propagate our network loopback prefix to the rest of the PEs, without creating full mesh of LSPs between the PEs and the RRs.

ihsan at kenanga# show label-switched-path to-self 

And CSPF kept mentioning empty route to the egress.

ihsan at kenanga# run show mpls lsp ingress name to-self extensive 
Ingress LSP: 17 sessions
  From:, State: Dn, ActiveRoute: 0, LSPname: to-self
  ActivePath: (none)
  LoadBalance: Random
  Encoding type: Packet, Switching type: Packet, GPID: IPv4
  Primary                    State: Dn, No-decrement-ttl
    Will be enqueued for recomputation in 10 second(s).
    1 Feb 11 06:08:58 CSPF failed: empty route
  Created: Sun Feb 11 06:08:55 2007
Total 1 displayed, Up 0, Down 1

I was also recommended to create the LSP to another RR and vice-versa (to get the prefix installed in inet.3) but I want to avoid that because in the event of the dest RR failing (we only have 2 RRs), then the NLRI will be marked inactive.

Any ideas anyone? 


> Hi Guys,
> What you are looking for is that in order for the l3bgp table to have active
> routes to pass to the other reflector clients the next hop for all PE¹s in the
> vpn global table (l3bgp table) must be a LSP learned route.  In the
> traditional sense this would require a full mesh of lsp¹s to the RR and a full
> mesh of lsp¹s to each PE.  A simple work around for this in order to not have
> lsp¹s going to the RR from each PE is to build a fake LSP on the RR to it¹s
> own loopback. Then install 0/0 under this LSP then their will be a 0/0 entry
> in inet.3 which will resolve all next-hops for the PE routes in the Global vpn
> table. Of course this will not actually forward traffic. The sole purpose of
> this 0/0 is to resolve routes and subsequently allow the RR to advertise the
> routes to other clients that are now active.
> I hope this clears things UP...
> Cheers,
> Gary Hauser 
> JNCIE #12, CCIE # 4489

More information about the juniper-nsp mailing list