MPLS-TE backup LSP protection behavior

Question

I'm trying to learn more about MPLS-TE, specifically how packets are routed once the TE LSP is set up.

Most of the examples I could find that include the MPLS label stack shown are examples for Fast Reroute, e.g. link/node protection. For example, if we have Router A -> Router B with the label 10 being pushed to get the packet from A -> B, with the protection LSP being A -> C -> B, when the A -> B link fails, we have another label, say 20, pushed on top of the label 10, so that traffic now goes through the protected LSP A -> C -> B.

Do all TE routes work this way? Does the original LSP path (w/o TE) have to be only one hop away? This is the part that confuses me - if it's multiple hops away, how would the head-end know which label to push to get to the tail-end after the packet exits the TE LSP?

For example, if my original LSP is A -> B -> C -> D -> E, and I have TE LSPs B -> F -> D and B -> G -> D, how would B know to push which LSP label onto the packet before it enters either TE LSPs so that when it gets to D, it can continue to E? There are 2 labels, the top one being the TE label, and the bottom one being the LSP label, right?

Answer 1

Do all TE routes work this way? Does the original LSP path (w/o TE) have to be only one hop away?

No the original LSP path can be multiple hops away.

MPLS TE is a very complicated subject, and I can hardly do justice to the nuances in a reasonable-sized answer. These are some of MPLS TE protection models for a primary LSP path along A -> B -> C -> D -> E:

 /------I-----\
/              \
A  -->  B  -->  C  -->  D  -->  E
 \     /\                      /
  \-H-/  \----------J---------/

• A-B Link protection model, as you described:
  • Option 1 (if A has a direct link to C): Backup LSP via A -> C -> B
  • Option 2: Backup LSP via A -> H -> B
• B-failure Node protection model: Backup LSP via A -> I -> C
• Path Protection model:
  • Primary TE LSP and Backup TE LSP share the same head and tail, but they must follow diverse physical paths
  • Primary LSP: B -> C -> D -> E
  • Backup LSP: B -> J -> E (both the primary and backup paths begin at B and end at E)

Please note that by default, MPLS TE tunnels are not advertised in the IGP, unless you configure a Forwarding Adjacency, which allows OSPF or ISIS to treat the tunnels as a routed link (although IGP adjacencies still don't form over the tunnel).

For example, if my original LSP is A -> B -> C -> D -> E, and I have TE LSPs B -> F -> D and B -> G -> D, how would B know to push which LSP label onto the packet before it enters either TE LSPs so that when it gets to D, it can continue to E?

It sounds like you're asking about node protection. This slide from Cisco Live 2011 BRKMPL-2104 does a good job explaining details of how labels are pushed. See the picture, below.

The normal MPLS TE LSP path is shown in yellow, R1 -> R2 -> R4 -> R6 -> R7, but R4 goes down. The original MPLS TE LSP signaled for R4 to swap label 22 (from R2) with label 36 as the LSP forwards through R6 to R7.

When R4 goes down, R2 (also called the PLR, point of local repair) now pushes two labels {25, 36} as it uses the green backup tunnel through R3. Since R3 is the penultimate hop of the backup tunnel it received an "implicit null" (i.e send no label) from R6 when the backup path was set up. Therefore, R3 strips label 25 off, and label 36 is still sent to R6.