XO - enable PCI devices for pass-through - should it work already or not yet?

MathieuRA

how many host do you have in your pool?

abudef

Hi @MathieuRA ,
There's only one host in the pool; all XCP-ng 8.3 patches installed, XO commit 6444f.

abudef

It behaves the same way in XOA 5.94.2

MathieuRA

So the error is "normal".
XOA tries to evacuate your host before restarting it, but your host's VMs have "nowhere to go" since you don't have another host.
Are you using XO from sources?

abudef

@MathieuRA I understand that, it's the same as having to reboot a host with running VMs, right? However, there we have an option called Force reboot. Or simply notify the user that the settings have been saved and the host needs to be rebooted, etc.

I tried this with both Xen Orchestra Appliance and Xen Orchestra from source (versions see above).

MathieuRA

@abudef As workaround, you can run an XO locally, shutdown all your VMs from your host and toggle your PCI from your local XO.
Otherwise it's also possible doing it with command line.

In the modal, if we simply change the buttons like this: [Reboot, Force Reboot, Cancel], instead of [Ok, Cancel] would that be sufficient?

Paolo

@MathieuRA said in XO - enable PCI devices for pass-through - should it work already or not yet?:

So the error is "normal".
XOA tries to evacuate your host before restarting it, but your host's VMs have "nowhere to go" since you don't have another host.
Are you using XO from sources?

IMHO If there isn't a host where to evacuate the running VMs, the reboot procedure should simply notify the user and shutdown the VMs.

abudef

@MathieuRA said in XO - enable PCI devices for pass-through - should it work already or not yet?:

In the modal, if we simply change the buttons like this: [Reboot, Force Reboot, Cancel], instead of [Ok, Cancel] would that be sufficient?

Yeah, that would be great. And it would also be handy to add a feedback in case of the "Reboot" option if running VMs can't be evacuated.

olivierlambert

That's the "force reboot" button we have. I agree that ideally we should catch that and offer the force alternative reboot for the user telling him "your VM will be rebooted".

abudef

That's exactly it, thank you