harvester: [Bug] Shutdown from inside the VM results in VM rebooting/UI stuck?

Hi team,

I’m currently seeing the following behavior for Linux and Windows VMs: when I try to shutdown the VM from the guest OS, Harvester restarts the VM automatically.

I tweaked the settings (see below) to reach a sort of middle ground: Harvester doesn’t restart the VM anymore, however the Harvester UI is not reflecting the “true” state of the VM and stays in a “pending” state.

Here is what I did and which results I have:

• Create/Modify a VM > edit the YAML config
• Replace the spec.running by spec.runStrategy

...
spec:
  runStrategy: RerunOnFailure
...

• Add the KubeVirt spec.spec.domain.features.acpi capability

...
spec:
  spec:
    domain:
      features:
        acpi:
          enabled: true
...

• Start/Connect to the VM
• Shutdown the VM

Results: While the VM will correctly shutdown, Harvester UI will “hang” on Starting with the message Virt-launcher pod is terminating. The kubectl get pods will show the VM pod as Completed and kubectl get vm will show the VM as Stopped:

Additional notes/hints: When the VM is stopped, if I click on Stop and check the VM config, I can see that spec.runStrategy has switched to halted.

I hope this is clear 😄

_Originally posted by @nunix in https://github.com/harvester/harvester/issues/574#issuecomment-1022457166_