Hi all,
the shellscript causing VMs to eventually stop when the host shuts
down/reboots has a window where it can loose ACPI shutdown events.
I have seen VMs not shutting down before they become ACPI aware,
causing the systemd magic to time out for a couple of minutes and
eventually killing the VM.
Steps to reproduce:
- start a new VM
- reboot the host while the VM is still booting up
What happens:
- the ACPI power-button event will get lost
- the systemd loop on the host will go and wait
- the VM will get killed hard eventually
What should happen (probably):
- the retry loop should inject an ACPI power-button-event with every
retry (5s?)
- the VM will eventually have ACPI support and pick up any of the
many button-events
- the VM will shut down on any of the many power-button events, if
there was enough time
I could try to come up with a patch, but i hope the description of the
problem is clear enough so someone else will try.
regards,
Henning