IGP-LDP and static route-LDP synchronization

With LDP, FECs learned from an interface do not necessarily link to that interface state. As long as the router that advertised the labels is reachable, the learned labels are stored in the incoming label map (ILM) table.

Although this feature gives LDP a lot of flexibility, it can also cause problems. For example, when an interface comes back up from a failure or from a shutdown state, the static routes bound to that interface are installed immediately. However, the LDP adjacency to the next hop might not be up, which means that the LDP SDP remains down. In this case, the MPLS traffic will be blackholed until the LDP adjacency comes up.

The same issue also applies to dynamic routes (OSPF and IS-IS).

To resolve this issue, the LDP synchronization timer enables synchronization of IGP or static routes to the LDP state.

With IGP, when a link is restored after a failure, IGP sets the link cost to infinity and advertises it. The value advertised in OSPF is 0xFFFF (65535). The value advertised in IS-IS regular metric is 0x3F (63) and in IS-IS wide-metric is 0xFFFFFE (16777214).

After IGP advertises the link cost, the LDP hello adjacency is brought up with the neighbor. The LDP synchronization timer is started by IGP from the time the LDP session to the neighbor is up over the interface. This synchronization timer allows time for the label-FEC bindings to be exchanged.

When the LDP synchronization timer expires, the link cost is restored and is readvertised. IGP will announce a new best next-hop and LDP will use it if the label binding for the neighbor FEC is available.

The preceding behavior is similar for static routes. If the static route is enabled for ldp-sync, the route is not enabled immediately after the interface to the next hop comes up. Routes are suppressed until the LDP adjacency with the neighbor comes up and the synchronization timer expires. The timer does not start until the LDP adjacency with the neighbor node is fully established.