RSVP Graceful Restart Operation

For RSVP graceful restart to function, the feature must be enabled on the global routing instance. RSVP graceful restart can be disabled at the protocol level (for RSVP alone) or at the global level for all protocols.

RSVP graceful restart requires the following of a restarting router and the router’s neighbors:

An object called Restart Cap that is sent in RSVP hello messages advertises a node’s restart capability. The neighboring node sends a Recover Label object to the restarting node to recover its forwarding state. This object is essentially the old label that the restarting node advertised before the node went down.

The following lists the RSVP graceful restart behaviors, which vary depending on the configuration and on which features are enabled:

You cannot explicitly configure values for the restart and recovery times.

Unlike other protocols, there is no way for RSVP to determine that it has completed a restart procedure, other than a fixed timeout. All RSVP graceful restart procedures are timer-based. A show rsvp version command might indicate that the restart is still in progress even if all RSVP sessions are back up and the routes are restored.