Today it was the second time I had a KVM/Qemu machine running Windows 2008 Server hanging subsequently stalling a reboot process on the Proxmox host.
The host is part of our production environment (cluster with 5 hosts) so it's only restartet after kernel upgrades from a cron job at 04:00 in the morning.
The server runs both this Win client and a VZ client. The VZ client was shutdown as expected but the KVM/Qemu Win client just hangs there waiting for me to log in and try to kill it. But I can't. From htop I tried to kill the parent process with a kill -15 but that darn thing seems to do nothing about it. It just sits there starring at me asking me what I think that was good for
Killing 1-4 of the 4 child processes does nothing either. All I get is "naa-nana-naa-naa, you can't kill me" bullying 
In another thread I read that I'm supposed to make sure that the Win client is set to allow Shutdown without logon which I have done with all of our Win clients (under Proxmox).
Is there a way to just kill a KVM/Qemu client until it's dead? I will go as far as saying that corrupting the qcow2 image is actually better (in our case) than the process is hanging and the reboot not completed since all our virtualized servers are load balanced.
Any advice is appreciated.
The host is part of our production environment (cluster with 5 hosts) so it's only restartet after kernel upgrades from a cron job at 04:00 in the morning.
The server runs both this Win client and a VZ client. The VZ client was shutdown as expected but the KVM/Qemu Win client just hangs there waiting for me to log in and try to kill it. But I can't. From htop I tried to kill the parent process with a kill -15 but that darn thing seems to do nothing about it. It just sits there starring at me asking me what I think that was good for


In another thread I read that I'm supposed to make sure that the Win client is set to allow Shutdown without logon which I have done with all of our Win clients (under Proxmox).
Is there a way to just kill a KVM/Qemu client until it's dead? I will go as far as saying that corrupting the qcow2 image is actually better (in our case) than the process is hanging and the reboot not completed since all our virtualized servers are load balanced.
Any advice is appreciated.