Active member in ClusterXL HA sends an ARP request for cluster VIP causing a temporary outage
The ClusterXL active member sends an ARP request for the cluster VIP.
The Standby member replies to this request with its MAC address.
This is causing an outage when using layer 3 switch, because it will learn the new ARP entry and will forward the traffic to the Standby member.
There are three known scenarios leading to this issue:
1. Multiple VSX clusters connected to the same physical layer 3 switch.
When cluster interface's network probing is triggered, VS starts sending ARP requests to multiple IP addresses in the network interface's VIP belongs to. While it never sends ARP requests to its own VIP, one or more of the target IP addressed may serve as a VIP of another cluster connected to the same switch.
Note that interface probing is not occurring normally and is only triggered when there is no Cluster Control traffic detected on a cluster interface.
2. Management HA running on two ClusterXL members
A process running on an active member may connect to a management process running to itself via VIP. This in turn may cause the active member to issue an ARP request for the VIP.
3. An administrator manually runs on the active member a command, script or process that connects to the cluster's VIP.
While theoretically there maybe other scenarios where the problematic ARP requests are sent, the fix provided in the hotfix is generic and solves the problem regardless of the trigger.