(automatic) force startup of single VM on single cluster node with missing quorum

[mscd]

Hello together,

I am playing around with a Proxmox cluster, where one node is a dedicated server hosted by Hetzner, two other nodes reside in my local home lab.
The connection between the dedicated node and my home lab is realized by a IPsec-VPN-tunnel, running on a virtualized OPNsense-VM on the Hetzner Proxmox node.

At the moment, I am facing the problem, that - for example - after a reboot of the Hetzner-server, this node is not able to start the OPNsense VM to get a reliable quorum through the other nodes in my home lab. I cloud solve this problem by temporarily reduce the quorum to 1 (on the Hetzner-node) ... but it would be much nicer to "allow" this Proxmox node to start the OPNsense VM (on local storage) in EVERY CASE, so that the IPsec-VPN-tunnel cloud be established afterwards, to get a synchronized cluster.

Is it possible to (automatically) force a start of a single VM on a Proxmox node with missing quorum?

Best regards,
mscd

 

[aaron]

You are using PVE in a way it was never intended to

If you want to start that VMwithout any quorum you can try to run it manually. The command to start the VM can be viewed with "qm showcmd <vmid>". If you execute it you will bypass the PVE tooling and the VM should start.

But again, having a cluster span from your home network to a node running in some datacenter will give you more headache down the line. If you added the 3rd node in the DC to have 3 votes in your cluster, have a look at the QDevice mechanism. You could install the corosync-qnetd service for exmaple on a Raspberri Pi and have a third vote in the local cluster at home.