Configuration Guide User guide

FastIron Configuration Guide 1347
53-1002494-02
BGP4 graceful restart
BGP4 Router A sends a Hold Time of 5 seconds and BGP4 Router B sends a Hold Time of 4
seconds, both routers use 4 seconds as the Hold Time for their BGP4 session. The default Hold
Time is 180 seconds. Generally, the Hold Time is configured to three times the value of the Keep
Alive Time.
If the Hold Time is 0, a BGP4 router assumes that its neighbor is alive regardless of how many
seconds pass between receipt of UPDATE or KEEPALIVE messages.
NOTIFICATION messages from BGP4 routers
When you close the router BGP4 session with a neighbor, or the router detects an error in a
message received from the neighbor, or an error occurs on the router, the router sends a
NOTIFICATION message to the neighbor. No further communication takes place between the BGP4
router that sent the NOTIFICATION and the neighbors that received the NOTIFICATION.
BGP4 graceful restart
BGP4 graceful restart is a high-availability routing feature that minimizes disruption in traffic
forwarding, diminishes route flapping, and provides continuous service during a system restart,
switchover, failover, or hitless OS upgrade. During such events, routes remain available between
devices. BGP4 graceful restart operates between a device and its peers, and must be configured
on each participating device.
Under normal operation, when a BGP4 device is restarted, the network is automatically
reconfigured. Routes available through the restarting device are deleted when the device goes
down, and are then rediscovered and added back to the routing tables when the device is back up
and running. In a network with devices that are regularly restarted, performance can degrade
significantly and the availability of network resources can be limited.
BGP4 graceful restart is enabled globally by default. A BGP4 graceful restart-enabled device
advertises this capability to establish peering relationships with other devices. When a restart
begins, neighbor devices mark all of the routes from the restarting device as stale, but continue to
use the routes for the length of time specified by the restart timer. After the device is restarted, it
begins to receive routing updates from the peers. When it receives the end-of-RIB marker that
indicates it has received all of the BGP4 route updates, it recomputes the new routes and replaces
the stale routes in the route map with the newly computed routes. If the device does not come back
up within the time configured for the purge timer, the stale routes are removed.
NOTE
BGP4 graceful restart is supported in FSX 800 and FSX 1600 Layer 3 switches with dual
management modules as well as in FCX switches in a stack . If the switch will function as a restart
helper device only, a secondary management module is not required.
This implementation of BGP4 graceful restart supports the Internet Draft-ietf-idr-restart-10.txt:
restart mechanism for BGP4
For details concerning configuration of BGP4 graceful restart, refer to “Configuring BGP4 graceful
restart” on page 1385.