This document explains reason for bgp session flap after RE switchover even though NSR is enabled.

Before switchover "show task replication" shows complete for all protocols.

After nonstop active routing switchover, BGP peer is flapped with event - RsyncIncomplete

Jul 31 05:42:03.901 2024 Lab_RE0 rpd[6567]: RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.10.10.1 (Internal AS 100) changed state from Established to Idle (event RsyncIncomplete) (instance master)

This issue can be seen when RE switchover is initiated while bgp replication is still "In progress".

If BGP is running with a large-scale routes or peers and NSR is enabled, after restarting routing on backup RE or rebooting backup RE, it may take time to sync info between master RE and backup RE.

During this time, if issuing the command “show task replication”, it might show BGP replication complete even before BGP NSR replication is complete. This allows system to be prematurely ready for switchover.

If RE switchover is performed while BGP is still syncing info to the backup RE, some un-synced peers may flap after switchover.

Task replication output shows bgp status "complete"

labroot@jtac-MX104-r006> show task replication 
        Stateful Replication: Enabled
        RE mode: Master

    Protocol                Synchronization Status
    BGP                     Complete              
    IS-IS                   Complete              
    PIM                     Complete              
    MPLS                    Complete              
    RSVP                    Complete              
    MVPN                    Complete    

The "complete" status in the output of "show task replication" only indicates that the socket replication has completed and the BGP synchronization can be "in progress".

Before attempting nonstop active routing switchover, check the output of "show bgp replication" to confirm that BGP routing table synchronization has completed on the backup Routing Engine.

labroot@jtac-MX104-r006> show bgp replication 
Synchronization master:
  Precision timer registration: NotRegistered
  Session state: Up, Since: 1:26:25
  Flaps: 1, Last flap reason: Socket replication deactivated
  Protocol state: Idle, Since: 1:22:09
  Synchronization state: Complete
  Number of peers waiting: AckWait: 0, SoWait: 0, Scheduled: 0
  Messages sent: Open 1, Establish 4, GrHelper 0, Update 5191, GrStaleLabel 0 Error 0, Complete 1
  Messages received: Open 1, Request 1 wildcard 0 targeted, EstablishAck 4, GrHelperAck 0, CompleteAck 1