MPLS IP - LDP Simple Lab

[Arthur Kant]

Hello, I have been doing alot of labbing and playing around with MPLS/LDP today with the following topology.

     A (PE) <-->  B (P) <--->  C (P) <--> D (PE)

I have have three seperate IGP AS/domains defined sharing the loopbacks between the below routers. 

ospf between A & B  

Eigrp between B & C

ospf between C & D

** No bgp **

LDP is working between A & B / B & C / C & D using loopbacks on each router.  My original goal was to have ip connectivity from A to D loopbacks w/o having an IGP end - to - end but I am failing.
My theory was that Router A would assign a local label to it's loopback, and then share that via label via LDP to Router B would share that across its LDP sesstion to C ..etc eventually populating the LFIB on D.

I know the answer is I need to have IGP connecting all loopbacks but I thought I could have seperate igp in the core between P routers and then the Label exchange do the rest of the work to glue everything together.

I know this is basic stuff here..

Thanks!

[Nagendra Kumar Nainar]

Hi Arthur,

I think you answered it :). Any router will not install the label entry in FIB/LFIB table if it doesnt have the prefix in RIB. IOW, for each prefix in RIB, a router will try to get the label from the label table and install in forwarding table.

In your topology, I think you dont redistribute between IGPs. so ideally loopback of A will not be known by C or D. From C point of view, it might receive the label information from B for loopback-A. But since it doesnt have any entry in routing table, it will not consider the label received from B.

You need to redistribute between IGPs to have it work.

-Nagendra

[Arthur Kant]

Thank you for the response.  So when I read about MPLS core (P) routers  architectures simply switching packets based on label.  That general description of packet movement can only be achieved in the core (assuming I am only using MPLS IP w/ LDP) by extending CE prefixes into the core using an IGP ?

Or does the description more refer to L3VPN scenarios and/or MPLS TE ?

[Nagendra Kumar Nainar]

Hi Arthur,

That general description of packet movement can only be achieved in the core (assuming I am only using MPLS IP w/ LDP) by extending CE prefixes into the core using an IGP ?

<Nagendra> No. CE prefixes are not required to be learnt by P routers. P router requires to know only about the other P and PE routers. In the below topology,

CE1-----PE1---P1---P2---PE2----CE2

P1 and P2 are required to know about PE1 and PE2. They dont need to know about CE1 or CE2. 

PE1 and PE2 can exchange CE details via BGP. So ideally PE1 will push 2 labels. top label is to reach PE2 (This is the label P1 or P2 will look into) and bottom label will be used by PE2 to identify CE2.

A detailed explanation about MPLS VPN is available in below link,

https://supportforums.cisco.com/video/11928951/ask-experts-webcast-introduction-mpls-vpn 

Or does the description more refer to L3VPN scenarios and/or MPLS TE ?

<Nagendra> In any of the MPLS services (L3VPN, L2VPN etc), P routers are only required to know about the path to reach PE routers.

-Nagendra