This role manages Xen VMs with systemd services.

For each configured VM a Xen VM configuration is generated. You can start and stop the VMs with systemd. Currently, iSCSI and NFS are completely supported as filesystem backends. For all other backends you manually need to install the system afterwards.

For each installed kernel, a dracut initrd will automatically created at /etc/xen/initrd.img-$VERSION (similar to /boot initrds for dom0). When starting a VM. /etc/xen/initrd.img will automatically be created as a symlink, pointing to the initrd of the currently running kernel.

• A Xen host machine running a Debian host system (you can use xen-setup to perform this task)
• Filesystem utilities for filesystems used by the guests
• python-netaddr on the executing machine

Start and stop the given VM. For example systemd start vm@stuvus-web01 will start the web01 server from the organisation stuvus.

Does a login or a logout for the iSCSI root disk for the given VM, this is only done when iSCSI is selected as root backend, otherwise systemd will ignore any request silently.

Mount or umount the root filesystem from the given VM. For example systemd start vm_mount@stuvus-web02 mounts the root filesystem of web01 @stuvus into /mnt/vms/stuvus-web01. The mount point is automated created and removed.

All NFS exports need to be accessible for the hypervisor as well as the respective VM. The exported path for the root filesystem must be of the following scheme: {{ xen_vman_nfsroot_base }}/<org name>/<vm name>-root .

¹ Variable is not required unless no default is given or other specified ² For further reference see man xl.cfg

This role currently supports a custom network script called vif-p2p. It is like (the official) vif-route, except that it configures the vif interface using a P2P configuration from dom0 to the guest.

vif-p2p supports IPv6. To use this IPv6 feature, you must yourself configure your VMs to use the link-local address fe80::2/64 and gateway fe80::1 on eth0. Those link-local addresses are fixed and are the same for all VMs.

That is, vif-p2p automatically configures the interface on the host like

ip address add 10.0.0.1/32 peer 10.0.0.10 dev vif10.0
ip address add fe80::1/64 dev vif10.0
ip route add 2001:DB8::10/128 via fe80::2 dev vif10.0

if 10.0.0.1 is the main IP of the hypervisor, 10.0.0.10 is the ip of the first VM interface and 2001:DB8::10 is the ipv6 of the first VM interface. (For IPv4, the kernel will automatically add the necessary route.)

In order to install vif-p2p you have to add the string "vif-p2p" to xen_vman_additional_network_scripts.

In order to install your own network script (that is not supplied by this role), you can add arbitrary paths to xen_vman_additional_network_scripts. Relative paths are relative to your files folder. Only the file name is preserved: For example, adding "xen_vman/vif-custom" to xen_vman_additional_network_scripts will instruct this role to copy files/xen_vman/vif-custom (on localhost) to /etc/xen/scripts/vif-custom (on the remote host).

For doing so, it is recommended to use a subfolder like xen_vman, as in the previous example. Otherwise your file name might collide with a supplied network script in a future version of this role. In the case of such a collision, the script supplied by this role takes precedence.

xen_vman_default_network_bridge is written to Xen's config file, while xen_vman_lazy_default_network_bridge is lazily evaluated in order to default the bridge of VM interfaces. Therefore you can access the VM variables from within xen_vman_lazy_default_network_bridge.

These variables can be specified for each VM. VMs are defined as elements of the xen_vman_vms list.

Each network interface may have the following options. All of them are optional.

Note that if you use xen_vman_default_vifname_prefix (and don't overwrite it using the vifname option), then .DEVID is appended to the vif name, whereby DEVID is the index of the interface. (Hence the name ..._prefix.) For example, when using xen_vman_default_vifname_prefix: "example", then the vifs are named example.0, example.1 and so on.

- hosts: hypervisor
  roles:
    - xen_vman
      xen_vman_vms:
        - name: testvm
          interfaces:
          - ip: 10.0.0.15
            ipv6: fd76:5eac:234:70c8::15
            mac: 00:14:22:01:23:45
            bridge: mybridge
            os:
              type: ubuntu
              version: 18.04

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

• Markus Mroch (Mr. Pi) markus.mroch@stuvus.uni-stuttgart.de