Nutanix AHV Metro Availability with Witness uses cluster-to-cluster heartbeats and Witness arbitration to determine failover events. Internal design notes state:
“In a Metro Availability scenario, if the primary cluster loses connectivity to the secondary cluster but remains operational, it continues serving I/O to the VMs. No automatic failover occurs unless the cluster is determined to be down.”
Also:
“I/O pauses only occur when the cluster cannot reach either the peer cluster or the Witness/Prism Central arbitration source.”
Applying this to the scenario:
Both metro clusters lose connectivity with each other.
The primary cluster also loses connectivity to Prism Central.
The recovery cluster remains connected to Prism Central.
Since the primary cluster cannot reach the recovery cluster nor Prism Central, the arbitration path is lost. This triggers the “I/O stall for safety” condition described as:
“I/O will be paused until the cluster regains quorum or arbitration to ensure against split-brain conditions.”
But because the primary clusteris still runningand is not declared failed (just isolated), VMs continue running and I/O is paused only as long as arbitration is unreachable.
Thus:
Option B is correct because it covers loss of connectivity to either peer or Prism Central.
Option C is correct because VMs do continue running; only I/O is paused to prevent inconsistency.
Automatic failover only happens when Witness declares the primary unreachable by quorum rules, which is not the case here since only one side can see the Witness.
