Caveats for a SummitStack
The following describes the hitless failover caveats for a SummitStack:
- All customer ports and the stacking links connected to the failed primary node will go down. In the recommended stack ring configuration, the stack becomes a daisy chain until the failed node restarts or is replaced.
- A brief traffic interruption (less than 50 milliseconds) can occur when the traffic on the ring is rerouted because the active topology becomes a daisy chain.
- Since the SummitStack can contain more than two master-capable nodes, it is possible to immediately elect a new backup node. If a new backup node is elected, when the original primary node restarts, it will become a standby node.
- To simulate the behavior of a chassis, a MAC address of one of the nodes is designated as the seed to form a stack MAC address. When a failover occurs, the SummitStack continues to be identified with this address.
- During an OSPF graceful restart, the SummitStack successfully restores the original link state database only if the OSPF network remains stable during the restart period. If the failed primary node provided interfaces to OSPF networks, the link state database restoration is prematurely terminated, and reconvergence occurs in the OSPF network due to the failover. See OSPF for a description of OSPF and the graceful restart function.
- During a BGP graceful restart, the SummitStack successfully restores the BGP routing table only if the BGP network remains stable during the restart period. If a receiving speaker detected the need for a routing change due to the failure of links on the failed primary node, it deletes any previous updates it received from the restarting speaker (the SummitStack) before the restart occurred. Consequently, reconvergence occurs in the BGP network due to the failover. See BGP for a description of BGP and its graceful restart function.