migration fails from 2.9.14 -> 8.0.3 / invalid bootorder: device 'scsi0' does not exist'

[rasanentimo]

We have been using this provider version 2.9.14 with proxmox 8.0.3 and we are testing the migration to version 3.0.1-rc6 with Proxmox 8.2.2.

We tried out following configuration which was working with with 2.9.14 / 8.0.3 (the only addition was the network id 0)

{
    "terraform": {
        "required_version": "1.7.5",
        "required_providers": {
            "proxmox": {
                "source": "Telmate/proxmox",
                "version": "3.0.1-rc6"
            }
        }
    },
    "resource": {
        "proxmox_vm_qemu": {
            "test-vm": {
                "provider": "proxmox.ch33",
                "name": "test-vm",
                "desc": "test-vm",
                "vmid": 3141,
                "target_node": "ch33",
                "full_clone": true,
                "clone": "ubuntu-2204-template",
                "boot": "order=scsi0;ide0",
                "scsihw": "virtio-scsi-single",
                "network": {
                    "id": 0,
                    "model": "virtio",
                    "bridge": "vmbr0",
                    "firewall": false,
                    "link_down": false
                },
                "cores": 4,
                "sockets": 1,
                "memory": 16384,
                "ssh_user": "root",
                "sshkeys": "ssh-rsa ",
                "os_type": "cloud-init",
                "ipconfig0": "ip=10.10.10.10/24,gw=10.10.10.1",
                "lifecycle": {
                    "ignore_changes": [
                        "qemu_os",
                        "ciuser",
                        "disk"
                    ]
                }
            }
        }
    }
}

and we are getting error

Error: error updating VM: 500 invalid bootorder: device 'scsi0' does not exist', error status: {"data":null}

If the vm is cloned via the GUI, it works fine.

Any ideas on how to troubleshoot this?

I tried to search for documentation for the migration path from 2.9.14 to 3.0.x but I couldn't find any.

[Tinyblargon]

@rasanentimo there is no disk configured, so it's likely trying to remove the disk, then scsi0 does not exist anymore.

[rasanentimo]

Thanks for the quick reply! So does it mean that the disk has to be created separately even if it's defined in the template (as it is in our case)?

[Tinyblargon]

Correct. This was done because there were many edge cases, whether the user intended for a disk to be there or for it to be removed.

[Tinyblargon]

@rasanentimo if you define the same disk in Terraform as in the template it will realize after migration that the disk is already there and will keep it.

The 2 important settings are the disks slot and capacity.