The vSphere HA agent on a host is in the Network Partitioned state. User intervention might be required to resolve this situation.
While the virtual machines running on the host continue to be monitored by the master hosts that are responsible for them, vSphere HA's ability to restart the virtual machines after a failure is affected. First, each master host has access to a subset of the hosts, so less failover capacity is available to each host. Second, vSphere HA might be unable to restart a Secondary VM after a failure (see Primary VM Remains in the Need Secondary State).
A host is reported as partitioned if both of the following conditions are met:
The vSphere HA master host to which vCenter Server is connected is unable to communicate with the host by using the management (or Virtual SAN) network, but is able to communicate with that host by using the heartbeat datastores that have been selected for it.
The host is not isolated.
A network partition can occur for a number of reasons including incorrect VLAN tagging, the failure of a physical NIC or switch, configuring a cluster with some hosts that use only IPv4 and others that use only IPv6, or the management networks for some hosts were moved to a different virtual switch without first putting the host into maintenance mode.
Resolve the networking problem that prevents the hosts from communicating by using the management networks.