Basic MPLS VPN needs loopbacks for PE routers

[vincehgov]

Why do the PE routers need to use loopbacks for iBGP peering in order for traffic to pass through.

I tried configuring with and without the loopbacks.  It doesn't work if I peer the two iBGP routers using physical interface IPs.

I get the error:

*Mar  1 02:44:48.159: %BGP-4-VPNV4NH_IF: Nexthop 10.3.4.4 may not be reachable from neigbor 10.2.3.2 - not a loopback

I want to know the technical reason as to why it wont work without it.  I tried looking at the ip cef table and it looks identical. 

Can someone enlighten me?

Thanks!

Vince

[vincehgov]

By the way, I can ping 10.3.4.4 from 10.2.3.2 just fine.  Those are the P facing interfaces of the PE routers.

[vincehgov]

Yikes! I should've searched first:

https://supportforums.cisco.com/thread/249438

[Harold Ritter]

Hi Vincent,

You are welcome ;-)

Have a great week end

[vincehgov]

Yeah thanks Harold.

To anyone else who stumbles upon this issue:

Since the next-hop address would be the PE's MPLS facing interface, the external MPLS label that is added targets the network segment that connects P to PE.  Since P is directly connected to that network, the previously hop would pop off the external label as per PHP behavior.

Therefor, peering with the loopbacks would cause a different external label to be added- one that will make it all the way to the PE.