True that suspend isn't ideal, but i only envisage this occurring during shutdown from UPS low battery condition, which should be quite a rare occurrence. In that situation, i would expect all clients to have disconnected from the DB as they would have already powered off. In other situations where I need to restart the host server for whatever reason (it's linux so that shoud also be quite rare), I could manually shutdown the VM first.
Now the issue with NT4 and Workstation is that the VM Shutdown function doesn't work. It triggers the shutdown correctly, but then goes into a permanent wait state where its expecting the VM to power off and that will never happen because NT4 doesn't support ACPI power functions. You have to kill the VM process to get out of it (there was another post on this forum about that).
So if Workstation handles the shutdown script by calling the script first and then calling it's internal shutdown handling, that won't work either because it will get stuck waiting for the OS to power itself off and that will never happen. That's why I thought if it were possible to somehow trigger the power off from within the VM, that would solve my problems. I can't really think of a clean solution to this yet. Other than some kind of poling script running on the host watching for some trigger inside the VM and powering off the VM when it gets it.
I've tried lots of posts I found about adding ACPI support to NT4 by using HALs found in recent service packs and making registry updates but none of them worked.
