Failover to Backup VSR-NRC

If the active VSR-NRC fails, the NRC-P detects it when the global CPROTO channel goes down.

The NRC-P uses a keep-alive timer of 60 seconds and a multiplier of 2.2 for a total keep-alive timeout of 132 seconds.

At keep-alive timeout, NRC-P determines that the channel is idle and closes it. NRC-P also closes the TOPO and PCEP CPROTO channels, if they are not already down.

Next, the NRC-P begins its CPROTO establishment cycle as detailed in Initial Establishment of Active/ Standby VSR-NRC Roles. If the failed VSR-NRC is the primary VSR-NRC (VSR-NRC-A in the primary site), three attempts are performed to bring the global CPROTO channel back up. If the global CPROTO channel is successfully restored, the primary VSR-NRC remains the target active VSR-NRC. If not, three attempts are performed to the secondary VSR-NRC (VSR-NRC-B) and so on.

Note: It can take up to 162 seconds for the NRC-P to switch to the secondary standby VSR-NRC. This includes the keep-alive idle time of 132 seconds plus up to 3 attempts of 10-second intervals to establish the global CPROTO channel, 3 unsuccessful attempts to primary VSR-NRC and one successful attempt to the secondary VSR-NRC. If the user reboots the primary VSR-NRC (VSR-NRC-A) while it is active, it may come back up faster and therefore it will remain the target active VSR-NRC and NRC-P does not switch to secondary VSR-NRC (VSR-NRC-B).

When the global CPROTO channel is up to the target active VSR-NRC, either the primary or secondary VSR-NRC, NRC-P sends a notify message requesting active role and providing the IP address for the mate VSR-NRC (wscIsActive=TRUE, mateAddr, matePort).

The target active VSR-NRC begins the transition to the active role by flapping the CPROTO sync session to the mate VSR-NRC, if not already down. Then it begins reconciling the local copy of the mate databases with its own network-learned databases. The common records only have a difference in the setting of the Delegate bit in the PCEP Report messages.

The following reconcile process is followed:

Common records of the local database are carried over with setting of the Delegate bit in the PCEP Report messages and with the LRID information from the mate database.
Records in the mate database which are not reconciled with the local database are deleted.
Records in the local database which are not reconciled with the mate database are always carried over.

Common records of the local database are preferred over those of the mate copy database except for the LRID. After the reconcile process is complete, the now newly active VSR-NRC destroys the mate copy database and acknowledges the NRC-P by sending a notify reply message (notifyReply=TRUE).

The newly active VSR-NRC begins the partial database synchronization procedures to NRC-P as described in TE-DB and LSP-DB Partial Synchronization. The newly active VSR-NRC stops accepting new records from its own TE-DB and LSP-DB until after it completes the reconcile between the mate copy databases and its local databases and completes the partial sync with NRC-P.

The newly active VSR-NRC notifies the local PCE process to set the PCEP overload to OFF and to start an overload timer, hard-coded to 10 minutes, for each PCEP session. At the receipt of the first PCEP redelegation from a PCC, VSR-NRC stops the timer for that PCC and sends a PCC ready message to NRC-P which can then begin sending update messages to that PCC. If the overload timer expires before receiving a PCEP redelegation message, the newly active VSR-NRC clears all delegations of the corresponding PCC toward NRC-P.

The newly active VSR-NRC attempts to establish a CPROTO channel to the mate VSR-NRC. After successfully establishing a channel, it begins the full or partial database synchronization procedures to the mate VSR-NRC following similar procedures as mentioned in TE-DB and LSP-DB Partial Synchronization.