A two-node cluster lacks the natural majority available in a conventional cluster containing three or more voting members. If communication between its two nodes is interrupted, each node could otherwise interpret the other node as failed and attempt to continue independently. That condition creates a split-brain risk because both sides could consider themselves authoritative.
The external Witness VM supplies the independent arbitration required for the two-node configuration. It is deployed in a separate failure domain and maintains communication with the cluster. During a node or communication failure, witness participation allows the cluster to distinguish the surviving authoritative side and avoid both nodes independently continuing cluster operations. Separating the witness from the two data- bearing nodes is critical because placing all voting components in the same failure domain would undermine the tie-breaking design.
Prism Element is the local cluster-management interface, while Prism Central provides centralized management across clusters. Neither substitutes for an independent quorum witness. Foundation Central supports centralized discovery and deployment activities and does not arbitrate two-node cluster ownership.
Nutanix documentation identifies the external Witness VM as part of the supported two-node ROBO configuration. References: Prism 7.5 - Two-Node Clusters and Witness for Two-Node Clusters . NHCF alignment: valid cluster configurations, availability, and failure handling.