VM migrations occur at vCenter server restart or enabling Proactive HA

Products

• VMware vCenter Server

Issue/Introduction

Symptoms:

There may be VM migrations after enabling Proactive HA on a cluster, or after vCenter server service restarts with Proactive HA enabled on a cluster. Some hosts in the cluster where migrations happen have a configuration issue of “xxx has unknown health state”.

Environment

VMware vCenter Server 8.0.x
VMware vCenter Server 7.0.x

Cause

After Proactive HA is enabled on a cluster, or after vCenter server service restarts with Proactive HA enabled on a cluster, the hosts in the cluster have an initial Proactive HA health state of “unknown” until the third party health provider posts the next health updates.
Distributed Resource Scheduler(DRS) treats hosts with “unknown” health state as “Quarantined” host, hence tries to avoid placing VMs on them if possible.
If DRS load balancing happens before all hosts in the cluster receive their health updates, VMs on hosts with “unknown” health state could be migrated to other hosts with “green” health state.

Resolution

This is a expected behavior as per the current design.

To avoid such VM migrations,

• set DRS to manual mode before restarting vCenter Server or enabling Proactive HA on a cluster.
• set it back to Fully Automated after all hosts receive at least one health update (host no longer has the “unknown health state” config issue).