vSphere HA initiated failover action to a single Virtual Machine even when VM monitoring disabled.

Products

• VMware vCenter Server
• VMware vSphere ESXi

Issue/Introduction

vSphere HA initiated failover action for a single Virtual Machine neither any host failure nor network failure.

Not related to vSphere HA feature "VM Monitoring".

In vCenter events, 

vSphere HA initiated a failover action on 1 virtual machines in cluster <<CLUSTER Name>> in datacenter <<Datacenter Name>>

Environment

vSphere HA enabled environment.

Cause

This issue can be observed on various reason but the followings are frequently reported.

• User initiated power action 'Hard stop' for the VM.
• VMX or VM monitor crashed.

When 'Hard Stop' operation is initiated, the following entry is recored in vCenter events,

Virtual machine <<VM Name>> termination requested

In hostd.log on ESXi on which the virtual machine was running, following entries is recorded.

Task Created : haTask-176-vim.VirtualMachine.terminate-<<TaskID>>

Attempting to terminate (SIGKILL) virtual machine with cartel id: <<CartelID>>

 

Resolution

This behavior is expected by design. 

vSphere HA monitors power state of each protected VMs and if detect unexpected power state change (no clean shutdown) initiate a failover action (restarting the VM) after short period of time for monitoring. (such as 30 seconds) No failover action is initiated when getting back to "Powered On" state in this period.

So consider to use "Shutdown Guest OS" or "Power Off" in usual case. not "Hard Stop".

What's "Hard Stop" opearation ?

You can use this operation only when you cannot power off a virtual machine or if a virtual machine becomes unresponsive. When you hard stop a virtual machine, all processes are ended and the virtual machine is powered off. By using this operation, you might lose all unsaved information.

If you use the Hard stop operation, vSphere HA restarts the virtual machine.