LDP FRR procedures

The LDP FEC resolution when LDP FRR is not enabled operates as follows. When LDP receives a FEC, label binding for a prefix, it will resolve it by checking if the exact prefix, or a longest match prefix when the aggregate-prefix-match option is enabled in LDP, exists in the routing table and is resolved against a next-hop which is an address belonging to the LDP peer which advertised the binding, as identified by its LSR-id. When the next-hop is no longer available, LDP deactivates the FEC and deprograms the NHLFE in the datapath. LDP will also immediately withdraw the labels it advertised for this FEC and deletes the ILM in the datapath unless the user configured the label-withdrawal-delay option to delay this operation. Traffic that is received while the ILM is still in the datapath is dropped. When routing computes and populates the routing table with a new next-hop for the prefix, LDP resolves again the FEC and programs the datapath accordingly.

When LDP FRR is enabled and an LFA backup next-hop exists for the FEC prefix in RTM, or for the longest prefix the FEC prefix matches to when aggregate-prefix-match option is enabled in LDP, LDP will resolve the FEC as above but will program the datapath with both a primary NHLFE and a backup NHLFE for each next-hop of the FEC.

In order perform a switchover to the backup NHLFE in the fast path, LDP follows the uniform FRR failover procedures which are also supported with RSVP FRR.

When any of the following events occurs, LDP instructs in the fast path the IOM to enable the backup NHLFE for each FEC next-hop impacted by this event. The IOM do that by simply flipping a single state bit associated with the failed interface or neighbor/next-hop:

  1. An LDP interface goes operationally down, or is admin shutdown. In this case, LDP sends a neighbor/next-hop down message to the IOM for each LDP peer it has adjacency with over this interface.

  2. An LDP session to a peer went down as the result of the Hello or Keep-Alive timer expiring over a specific interface. In this case, LDP sends a neighbor/next-hop down message to the IOM for this LDP peer only.

  3. The TCP connection used by a link LDP session to a peer went down, due say to next-hop tracking of the LDP transport address in RTM, which brings down the LDP session. In this case, LDP sends a neighbor/next-hop down message to the IOM for this LDP peer only.

  4. A BFD session, enabled on a T-LDP session to a peer, times-out and as a result the link LDP session to the same peer and which uses the same TCP connection as the T-LDP session goes also down. In this case, LDP sends a neighbor/next-hop down message to the IOM for this LDP peer only.

  5. A BFD session enabled on the LDP interface to a directly connected peer, times-out and brings down the link LDP session to this peer. In this case, LDP sends a neighbor/next-hop down message to the IOM for this LDP peer only. BFD support on LDP interfaces is a new feature introduced for faster tracking of link LDP peers.

The tunnel-down-dump-time option or the label-withdrawal-delay option, when enabled, does not cause the corresponding timer to be activated for a FEC as long as a backup NHLFE is still available.