Proxmox Virtual Environment: Difference between revisions

From NixOS Wiki
imported>Bradfitz
m typo fixes
imported>Lindenk
m fixed typo
Line 14: Line 14:
It is possible to generate generic qcow2 images and attach them to VMs with <code>qm importdisk</code> as shown [https://pve.proxmox.com/wiki/Migration_of_servers_to_Proxmox_VE#Importing_to_Proxmox_VE here].
It is possible to generate generic qcow2 images and attach them to VMs with <code>qm importdisk</code> as shown [https://pve.proxmox.com/wiki/Migration_of_servers_to_Proxmox_VE#Importing_to_Proxmox_VE here].


A better option is to generate a VMA image that can be imported as a VM on proxmox. With this method, many VM configuration options such as CPU, memory, netowrk interfaces, and serial terminals can be specified in nix instead of manually setting them on the proxmox UI.
A better option is to generate a VMA image that can be imported as a VM on proxmox. With this method, many VM configuration options such as CPU, memory, network interfaces, and serial terminals can be specified in nix instead of manually setting them on the proxmox UI.


=== Generating VMA ===
=== Generating VMA ===

Revision as of 15:48, 11 February 2024

Proxmox Virtual Environment - shortened PVE - (wikipedia:en:Proxmox Virtual Environment) is a platform for containerization and virtualization. PVE can manage a so called "data center" as a cluster of machines and storage. (It supports file systems like ZFS and Ceph.) Mostly it can be used with a WUI. It is open source and is based on Debian GNU/Linux (with a customized kernel of Ubuntu).

PVE uses

  • #LXC for containerization and
  • #KVM for virtualization.

NixOS runs on both.

The instructions should work for PVE 7.2 and later with NixOS 22.05 and later.

KVM

It is possible to generate generic qcow2 images and attach them to VMs with qm importdisk as shown here.

A better option is to generate a VMA image that can be imported as a VM on proxmox. With this method, many VM configuration options such as CPU, memory, network interfaces, and serial terminals can be specified in nix instead of manually setting them on the proxmox UI.

Generating VMA

The first run will take some time, as a patched version of qemu with support for the VMA format needs to be built

nix run github:nix-community/nixos-generators -- --format proxmox

Pass additional nix configuration to the template with --configuration filename.nix. In addition to NixOS module options, proxmox-specific options present in nixos/modules/virtualisation/proxmox-image.nix can be used to set core, memory, disk and other VM hardware options.

Deploying on proxmox

The generated vma.zst file can be copied to /var/lib/vz/dump/ (or any other configured VM dump storage path). A new VM can be spun up from it either using the GUI or the CLI:

qmrestore /var/lib/vz/dump/vzdump-qemu-nixos-21.11.git.d41882c7b98M.vma.zst <vmid> --unique true

note: the MAC address of net0 defaults to 00:00:00:00:00:00. This must either be overridden through proxmox.qemuConf.net0, or the unique attribute must be set to true when importing the image on proxmox.

By default, the generated image is set up to expose a serial terminal interface for ease of access.

root@proxmox-server:~# qm start <vmid>
root@proxmox-server:~# qm terminal <vmid>
starting serial terminal on interface serial0 (press Ctrl+O to exit)

<<< NixOS Stage 1 >>>

loading module dm_mod...
running udev...
Starting version 249.4
.
.
.
[  OK  ] Reached target Multi-User System.


<<< Welcome to NixOS 21.11.git.d41882c7b98M (x86_64) - ttyS0 >>>

Run 'nixos-help' for the NixOS manual.

nixos login: root (automatic login)


[root@nixos:~]#

Network configuration

Cloud-init can be enabled with

services.cloud-init.network.enable = true;

This will enable systemd-networkd, allowing cloud-init to set up network interfaces on boot.

LXC

Generating LXC template

nix run github:nix-community/nixos-generators -- --format proxmox-lxc

Privileged LXCs

While it’s not necessary, proxmoxLXC.privileged can be set to true to enable the DebugFS mount in privileged LXCs. If enabled on unprivileged LXCs, this will fail to mount.

Network configuration

The proxmox LXC template uses systemd-networkd by default to allow network configuration by proxmox. proxmoxLXC.manageNetwork can be set to true to disable this.

deploying on proxmox

Copy the tarball to proxmox, then create a new LXC with this template through the web UI or the CLI. The “nesting” feature needs to be enabled. Newer versions of proxmox will have it enabled by default.

As of now, not all of the configuration options on the web UI work for proxmox LXCs. Network configuration and adding SSH keys to root user work, while setting a password for the root user and setting hostname don’t.

It is suggested to set a root password within the container on first boot.

The template built above without any options does not come with /etc/nixos/configuration.nix. A minimal working example is presented below. Be sure to run nix-channel --update, reboot the container running before nixos-rebuild switch.

{ pkgs, modulesPath, ... }:

{
  imports = [
    (modulesPath + "/virtualisation/proxmox-lxc.nix")
  ];

  environment.systemPackages = [
    pkgs.vim
  ];
}

LXC Console

You may need to set the Console Mode option to /dev/console (instead of the default of "tty") in order to make the console shell work.

LXC See also

Name

Proxmox Virtual Environment is also called

short Proxmox VE,
shortened PVE,
just Proxmox.

Proxmox is the firm of the company Proxmox Server Solutions GmbH. Besides Proxmox Virtual Environment (PVE)[1] there are other products called Proxmox Backup Server (PBS)[2] and Proxmox Mail Gateway (PMG)[3].

References