MPLS FRR

[Mohammed Alaradi]

Dears,

when we deploy MPLS FRR link/node protection, at any of these cases the PLR will swap the the comming traffic label with either

Nhop label or NNhop lablel (bypass tunnel) and push on top of transport label the FRR label. so the (transport label) is already known to

the (MP) before any failure.

My confuse, when the link/node flap that cause FRR to be on action, what will keep the MP still holding/installing the protected LSP Label while the link/node that is connected to is already down.. in that case there is no meaning of FRR, especialy we used BFD to detect the failure

and RSVP could torn down the session directly with it's orginal label.

any one could clear this point.

Mohamed.

[Peter Paluch]

Mohamed,

what will keep the MP still holding/installing the protected LSP Label

There is nothing special about the label assigned to the particular FEC by the MP router. This label does not have any special significance to the whole FRR mechanism.

If there was no failure in the network, MP would be receiving packets for this FEC labeled with this label from its directly connected neighbors. After a failure in network occurs, the PLR still has a knowledge of this label and it imposes it onto the packet, but because now the packet has to go through a precomputed detour, yet another label will be pushed onto the packet that helps to carry the packet over the repair path. The last router before the MP will pop this "repair path" label and MP will receive the packet with the label for the FEC just like if nothing happened at all.

Would this help a bit? Please feel welcome to ask further!

Best regards,
Peter

[Mohammed Alaradi]

Hello Peter,

Your answer is how PLR reroute and encapsulate the packet for FRR.

my Q: "PLR still has a knowledge of this label and it imposes it onto the packet",

but the link/Node flap and the label that is already encapsulated to the stack for protected LSP by PLR should be drop when reach MP because the link advertise this label flap, else the network will have black holing labels while network failure, but for FRR, what make the protected LSP label install  swap entry in FIB @MP and not drop the protected LSP label.

this my confuse and looking for any clarification.

Mohamed.

[Peter Paluch]

Hi Mohamed,

but the link/Node flap and the label that is already encapsulated to the stack for protected LSP by PLR should be drop when reach MP because the link advertise this label flap

No, no, that is not the case. Routers assign labels to destinations (or FECs) regardless of the incoming interface. Even though an interface may have failed on an MP, this does not invalidate the labels that have been assigned by the MP to individual destinations. In other words, if the destination X has been assigned the label Lx by an MP, then a packet labeled with Lx can arrive through any interface on the MP, and it will still be accepted and forwarded along the path toward the destination X. This is what makes the FRR work - after the primary path fails, a backup path kicks into action on PLR, but as long as the MP gets the packet with the original label, no matter through which interface, this packet will be accepted by MP and forwarded along its proper path.

There was a mode of MPLS operation where labels were assigned on a per-interface basis, or in other words, the router was using a per-interface label space. In this mode, each incoming interface maintained its own label mappings, and consequently, its own LFIB. Label Lx advertised from interface 1 was entirely independent of the same label Lx advertised from interface 2, and could refer to an entirely different destination. These per-interface label spaces were intended to facilitate interoperation between MPLS and ATM in cell mode, as ATM operated with its VPI and VCI identifiers in a per-interface mode. However, in MPLS packet mode which is the only one widely used nowadays, per-interface label spaces are not used. Instead, all Cisco routers default to per-platform label space where a label is assigned just to a FEC, ignoring the incoming interface entirely. In fact, for FRR, the use of per-platform label space is a requirement - exactly to avoid the problem you're discussing here.

As always, you're welcome to ask further!

Best regards,
Peter

[Mohammed Alaradi]

Hello Peter, thanks for your continue support.

The Protected LSP path is a collection of RSVP sessions across multiple LSR over the HE-->TE path, each LSR base on the RSVP session (LSP) going to assign a label and create PSB and RSB for that session + Label (cross-connect) with incoming and outgoing label.

Agree, the Router will accept the label as far the platform mode is used from any interface, however FRR is valid (IF) the router (MP)  is already installed the label swap for the "protected LSP" label and not drop it, there are many cases of when the router will drop the label, one of them is the session is timeout.

The "KEY" is PLR going to (Refresh) the MP with "Orginal LSP session" PATH message and the MP in return going to "Refresh" the PLR with "original RESV Message" which contains the Original label that later will be pushed as transport by PLR,, thus the session will not timeout and the MP will keep installing the original label in LFIB for swap operation, your kind clarification is valid onward of this step where the router already install the swap entry and will accept the label from any interface base on Platform-wide mode.

so the key point is that the RSVP message Refresh between PLR <--> MP for Protected LSP through the frr tunnel is what keep MP to keep installing the original swap label and not drop it when the failure is kick off.

Mohamed