The Active function is assumed by both Edges in the event that the HA link is severed or if the Active and Standby Edges are unable to communicate with one another. Therefore, both Edges begin responding to ARP requests on their respective LAN interfaces as a consequence of this. It is possible that spanning tree loops will occur on the local area network (LAN) as a consequence of this redirected LAN traffic to both Edges.
Switches are typically responsible for implementing the Spanning Tree Protocol in order to avoid loops in the network. In this scenario, the switch would prevent traffic from reaching either one of the Edges or both of them. Nevertheless, doing so would result in a complete shutdown of the traffic that passes via the Edge pair.
In an Enhanced HA implementation, split-brain detection requires communication to the Primary Gateway (i.e., there is no Layer 2 Switch attached to the Edge’s WAN interfaces).
Identification and Prevention of Split-Brain
This part talks about how to find and stop a split-brain state in an Edge setup with a high availability topology.
In a high availability deployment, there are two ways to find and stop a split-brain situation, which happens when both HA Edges become active.
When the HA heart rate link between the devices is lost, the first method sends layer 2 broadcast heartbeats between the two HA Edges. It sends a layer 2 broadcast (EtherType 0x9999) heartbeat on all of its WAN ports from the Active Edge to try to find the Standby Edge in the broadcast network. When this packet reaches the Standby Edge, it sees it as a message to stay in the Standby state. This method is used in a Legacy High Availability setup where both HA Edges’ WAN ports are linked to the same layer 2 switch.
