If you run sudo systemctl reboot inside a VM deployed with this driver, the task is failed and restarts (if configured with an appropriate restart block). But the restarted task is completely new. The original domain is destroyed and the new domain gets a new DHCP lease, if you've thin-provisioned a disk that'll be replaced, etc.
Maybe this is ok for some workloads, but it feels like surprising behavior for VMs (vs containers, which we expect to be ephemeral). Losing the disk is unfortunate though.
If you run
sudo systemctl reboot
inside a VM deployed with this driver, the task is failed and restarts (if configured with an appropriaterestart
block). But the restarted task is completely new. The original domain is destroyed and the new domain gets a new DHCP lease, if you've thin-provisioned a disk that'll be replaced, etc.Maybe this is ok for some workloads, but it feels like surprising behavior for VMs (vs containers, which we expect to be ephemeral). Losing the disk is unfortunate though.