OSPF resiliency and graceful restart

Additional reference information

• Routers' neighbors must cooperate as helpers during a graceful restart.

• All neighbors must support graceful restart.

• Graceful restart does not occur at initial router startup.

• Simultaneous graceful restart sessions on multiple routers in a single segment are not supported.

• OSPFv3 neighbor and database information are not checkpointed.

• OSPFv3 reestablishes adjacencies after a restart and requires consistent pre- and post-restart configurations.

• IPv6 FIB tables remain initially unchanged, but routes are eventually marked stale for convergence.

• OSPFv3 must send hellos within the dead interval after restart (default is 40 seconds), or adjacencies will drop.

• The feature manages OSPFv3 route purge timers and associated grace link-scope LSAs.

• Adjacency creation times are stored for post-restart uptime reporting.

• Strict-mode and non-strict-mode are incompatible; mixing them prevents OSPF neighbor relationships from forming.

• Both BFD neighbors must run IOS XR images supporting strict-mode. Mixing strict and non-strict is not supported.

• If BFD sessions are established by other clients, a neighbor relationship may still form.

• The multicast-intact feature enables multicast routing (PIM) when IGP shortcuts are configured and active.

• Both OSPFv2 and IS-IS support multicast-intact.

• When enabled, IGP provides alternate multicast-intact next hops for use by PIM.

• These next hops do not include IGP shortcuts and are exclusively for PIM.

• They are not published to the FIB or used for unicast routing.

• All link-state learned IPv4 destinations are published with multicast-intact next hops to the RIB.

• OSPF applies the max-paths limit separately to native and multicast-intact next hops.

• Due to BFD dependency, OSPF in strict-mode may have delayed neighbor establishment and full adjacency.

• Warm standby enables a standby OSPF process on the backup Route Processor (RP) to initialize before failure, reducing switchover downtime.

• Warm standby is always enabled for OSPFv3, supporting high availability.

• Nonstop routing (NSR) makes process restarts or upgrades invisible to peer routers, eliminating the need for traditional graceful restart protocol extensions.

• To restore OSPFv3 after a shutdown triggered by the protocol shutdown command, use no protocol shutdown. If shutdown was due to critical memory, wait for memory recovery and manually restart OSPFv3.

• Upon restoration, OSPFv3 brings up all interfaces with hello packets, rebuilds and advertises local LSAs, responds to neighbor messages, and installs learned routes in the RIB.

Events on OSPFv3 restoration:

1. All OSPFv3 interfaces are activated and exchange database information.

2. Local router and link LSAs are rebuilt and advertised.

3. The router replies to incoming OSPFv3 control messages.

4. Learned routes are installed in the RIB.

In shutdown mode, OSPFv3 disables itself by flushing self-originated LSAs, taking down interfaces, and clearing the LSDB. Shutdown may be manual (protocol shutdown) or due to memory exhaustion.

• All local LSAs are flushed.

• Incomplete adjacencies are dropped.

• After a delay, empty hello packets are sent, ceasing after the dead interval.

• The LSDB and OSPFv3 RIB routes are purged after the dead interval.

• The router ignores OSPF packets while shut down.

Helper Mode: Enabled by default. A router enters helper mode when it receives a grace LSA Type 11 from a neighbor in graceful restart.

• If helper mode is disabled with graceful-restart helper disable, LSA packets are dropped.

• The helper function operates for a set duration based on the grace LSA, stopping upon successful restart or expiration.

• The dead timer is ignored in helper mode.

• The helper role ends upon reestablishing full adjacency or timer expiry.

Restart Mode: OSPFv3 checks if a restart is needed on process start, based on prior enablement (not on first boot). When enabled, it adjusts the RIB purge timer.

• During restart, routes are not populated in the RIB; full adjacencies from before are restored.

• Repeated restarts are prevented by setting a minimum interval with graceful-restart interval.

• Restart duration can be set with graceful-restart lifetime; a grace LSA Type 11 is sent for neighbor notification and helper mode entry.

• Preserves data plane capability during:

  • RP failure and switchover

  • Planned or unplanned process restarts (such as upgrades or process crashes)

  • Critical memory events triggering autonomous shutdown

• Enables nonstop forwarding on established routes during OSPFv3 restart, improving IPv6 forwarding high availability.

• NSF enables OSPF routers to continue forwarding packets during process restarts or RP failover.

• Peer devices do not experience routing flaps.

• RPs synchronize FIBs and maintain neighbor and link-state information through restart events.

• After restart, neighbor relationships and OSPF databases are rebuilt, stale routes are cleared, and the RIB/FIB are updated once full synchronization completes.