Seamless MPLS Problem

[Joe Bee]

Today I was studying the seamless MPLS approach from this page:
https://www.cisco.com/c/en/us/support/docs/multiprotocol-label-switching-mpls/mpls/116127-configure-technology-00.html

But there is a point where I really struggle.
See the picture below.

How can I separate the LDP path? I can activate the LDP with "mpls ip" which also starts LDP.
This automatically forms neighbours and brings up a complete LDP based labeled path. No separation.
Unfortunately the IP, MPLS and LDP settings are not part of the examples. Can somebody help?

[Harold Ritter · Accepted solution]

Each domain (aggregation, core, aggregation) runs its own IGP and doesn't share LDP between each other. To do this, you do not enable LDP on the interface(s) between the different domains. You rather use BGP ipv4 label unicast to signal the LSP.

Regards,

[Joe Bee]

Hello Harold, thanks for your answer.
But we have this Edge Router from the picture above.
It has one interface to the left with LDP active, and one Interface to the right with LDP active.
How can I "disconnect" the LDP protocol on one physical router?

For the IGP this is simple. Just using different processes ore different IGPs, but how can this be achieved for LDP?
Hello Harold, thanks for your answer.
But we have this Edge Router from the picture obove. It has one interface to the left with LDP active and one Interface to the right with LDP active. How can I "disconnect" the LDP protocol on one physical router?

Fort the IGP this is simple by using diffrent processes ore diffrent IGPs, but how can this be achieved for LDP?

[Harold Ritter · Accepted solution]

In your scenario, you do not need to disable LDP on the edge router core interface. Just the fact of running a different IGP in the aggregation and core will disconnect these networks. This is where BGP label unicast provides the end to end glue.

Regards,

[Joe Bee]

Thank you Harold.

I think I got it now. Sometimes you didn't see the forest with all these trees in your way. ;)
In this example they redistribute just the loopbacks of RR1 in ospf2 and RR2 in ospf3.
LDP neighbor ship, even if it is established between RR1 and PE1, will not build up any labeled path for prefixes outside ospf 2
That's the trick. Aad then comes the Seamless MPLS to provide you with the full labeled path.