vSphere HA failover alerts triggered during staggered network recovery from host isolation

Products

• VMware vSphere ESXi

Issue/Introduction

Following a cluster-wide network outage and subsequent HA network isolation, vCenter may report HA failover failures for VMs that are still powered on.

This occurs when some hosts rejoin the cluster faster than others, leading the master to attempt restarts for VMs.

 

Environment

VMware vSphere ESXi

Cause

All ESXi hosts in one cluster entered an isolated state after a network failure.

Except one ESXi host, other ESXi hosts in this cluster rejoined the HA cluster and elected a master. As one ESXi still remained partitioned, the HA master node cannot determine the power state of VMs on this ESXi, leaving these VMs power states as unknown.

HA master node initiates a failover for these VMs, which fails because the VMs are still running and the file locks (.vmx) cannot be acquired by other hosts.

Resolution

This is a transient condition. Once the remaining host rejoins the cluster, the HA master will synchronize the power states. The failover alerts can be safely ignored and cleared if the VMs are confirmed to be running on their original host.