When a BGP Graceful Restart-capable router loses connectivity, it performs the following actions as the restarting router:
The restarting router establishes BGP sessions with other routers and relearns the BGP routes from other routers that are
also capable of Graceful Restart. The restarting router waits to receive updates from the neighboring routers. When the neighboring
routers send end-of-Routing Information Base (RIB) markers to indicate that they are done sending updates, the restarting
router starts sending its own updates.
The restarting router accesses the checkpoint database to find the label that was assigned for each prefix. If it finds the
label, it advertises it to the neighboring router. If it does not find the label, it allocates a new label and advertises
The restarting router removes any stale prefixes after a timer for stale entries expires.
A BGP Graceful Restart-capable peer router performs the following actions when it encounters a restarting router:
The peer router sends all the routing updates to the restarting router. When it has finished sending updates, the peer router
sends an end-of-RIB marker to the restarting router.
The peer router does not immediately remove the BGP routes learned from the restarting router from its BGP routing table.
As it learns the prefixes from the restarting router, the peer refreshes the stale routes if the new prefix and label information
matches the old information.
If a router is not configured for the NSF/SSO--MPLS VPN feature and it attempts to establish a BGP session with a router
that is configured with the NSF/SSO--MPLS VPN feature, the two routers create a normal BGP session but do not have the ability
to perform the NSF/SSO--MPLS VPN feature.