mbprtpmnr / vms.nix

NixOS module for QEMU VM's using a similar interface to nixos-containers.

vms.nix

A NixOS module that provides QEMU based virtual machines using a similar configuration interface to the containers NixOS options.

Usage

Before proceeding you must first import this module. Simply clone this repository and place it under /etc/nixos/ then in your /etc/nixos/configuration.nix add the following imports:

{
  imports = [
    ./vms.nix
  ];
}

Enabling

Everything this module does can be enabled or disabled with the boot.enableVirtualMachines, so the first step is to set that to true somewhere in your NixOS configuration.

{
  boot.enableVirtualMachines = true;
}

If we later set this option to false all virtual machines will be stopped and their services removed, although any data will persist.

First Virtual Machine

Let's test the waters by starting a dead simple VM that we can remotely connect to:

{
  boot.enableVirtualMachines = true;

  virtualMachines.guest = {
    autoStart = true;

    config = {
      system.stateVersion = "19.03";
      services.openssh.enable = true;
      users.users.root.openssh.authorizedKeys.keyFiles = [
        /root/.ssh/authorized_keys
      ];
    };

    qemuSwitches = [
      "nic user,hostfwd=tcp::10022-:22"
    ];
  };
}

Here we have defined a virtual machine called guest that runs openssh, allowing in to root anyone that has been given remote access to the root user of the NixOS host machine via the /root/.ssh/authorized_keys files.

Finally we add an extra switch to QEMU to be used when the VM starts that sets up network forwarding from localhost:10022 on the host machine to port 22 in the guest machine.

Now we can run nixos-rebuild switch and the VM will come online ready for connection. First we will check the VM service's status where we should be able to see the boot log.

systemctl status vm@guest

All virtualMachines are managed via a systemd service named vm@$VM_NAME. We can also stop, start, or restart this service.

Before a machine starts (other than its first time) it will attempt to migrate the backing store of the disk image, this means that the config is deployed onto a disk image in your /nix/store but any changes to the VM once it is running are stored on the root disk image at /var/lib/vms/$VM_NAME.$VM_ROOT_IMAGE_FORMAT so if we wanted to erase any state that has built up for the guest VM I can delete /var/lib/vms/guest.qcow2. This path can be changed via the virtualMachines.guest.rootImagePath option.

Let's get on with it and get in there, assuming you have a key in /root/.ssh/authorized_keys:

ssh -p 10022 root@localhost