Virtual Machines which are part of an affinity rule may not restart during a HA event

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms:

• Virtual Machines are part of a DRS affinity rule to keep virtual machines together.
• vSphere HA is enabled on the cluster.
• There are sufficient failover resources or a dedicated failover host in the cluster to restart all virtual machines which are part of the affinity rule.
• A user initiates an Enter Maintenance Mode task on the ESXi host which has just failed or is in a not responding state.
• Virtual machines which are part of DRS affinity rules are not restarted by vSphere HA but other virtual machines on the failed host are restarted successfully.
• You observe warnings such as the following in the vCenter Server events tab of the failed ESXi host:

Insufficient resources to fail over VirtualMachine01 in Cluster01 that resides in Datacenter01. vSphere HA will retry the fail over when enough resources are available. Reason: The anti-affinity rule set on your virtual machine prevents the placement of virtual disks

Environment

VMware vSphere ESXi 6.7
VMware vSphere 7.0.x

Cause

If the vCenter Server has not yet marked a failed ESXi host as not responding, the virtual machines on it may still have a status of powered on from the vCenter Server perspective. vSphere HA will keep retrying to fail over the virtual machines until it gets a successful placement, however if an affined virtual machine is already seen as powered on then DRS will attempt to place its partner virtual machine on the failed ESXi host.

Resolution

This issue is resolved with VMware vCenter Server 7.0 U3 and above.

Workaround:

The following workaround may be implemented until a fix is available:

• Do not invoke an enter maintenance mode task on a host which is in NotResponding mode or has just failed.