Cisco Cisco IOS Software Release 12.2(27)SBC

When a router that is capable of BGP Graceful Restart loses connectivity, the following happens to the 
restarting router:

The 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 it. 

The restarting router removes any stale prefixes after a timer for stale entries expires. 

When a peer router that is capable of BGP Graceful Restart encounters a restarting router, it does the 
following:

The peer router sends all of 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.

This section contains the following procedures:

 (required)

 (required) 

 (optional)

Perform this task to configure NSF support for basic VPNs.

Route Processors must be configured for SSO. See the

 feature module for more 

information.