Olive 2022

Edge-native virtual desktop application that uses the Sinfonia framework to discover a nearby cloudlet to run the virtual machine.

Virtual machine images from Olivearchive are converted from their original vmnetx package format to a containerDisk that can be executed with KubeVirt. The containerDisk images can be pushed into a private Docker registry.

Installation

Olive2022 depends on an available VNC client. I've found that virt-viewer from libvirt generally does a good job. On a Debian/Ubuntu system this can be installed with.

sudo apt install virt-viewer

It is best to manage the installation of olive2022 in a separate virtualenv with pipx.

python3 -m pip install --user pipx
python3 -m pipx ensurepath

Once these dependencies are in place, installation should be fairly straightforward, even specifying a python version should only be necessary if the system default happens to be older than Python-3.7.

pipx install olive2022

If installation fails at any point, there are various troubleshooting tips at the end of this document.

Usage

olive2022 install creates a .desktop file to declare a handler for vmnetx+https URLs.

When you then 'Launch' a virtual machine from the Olivearchive website, the handler will execute olive2022 launch with the VMNetX URL for the virtual machine image.

Internals

olive2022 launch hashes the VMNetX URL to a Sinfonia UUID, and uses sinfonia-tier3 to request the relevant backend to be started on a nearby cloudlet. When deployment has started, sinfonia-tier3 will create a local wireguard tunnel endpoint and runs olive2022 stage2 which waits for the deployment to complete by probing if the VNC endpoint has become accessible. It will then try to run remote-viewer (from the virt-viewer package), gvncviewer, or vncviewer.

Converting VMNetX packages

olive2022 convert will take a VMNetX URL, download the vmnetx format package file and convert it to a containerDisk image and associated Sinfonia recipe. The Docker registry to push the containerDisk image to can be set with the OLIVE2022_REGISTRY environment variable. If it is a private repository, the necessary pull credentials to add to the recipe can be specified with OLIVE2022_CREDENTIALS=<username>:<access_token>.

Installation troubleshooting

/usr/bin/python3: No module named pip

Pip is not installed on your system. On Debian/Ubuntu systems, to reduce the chance of interfering with packaged Python modules, the default Python installation does not install pip and even disables the python3 -m ensurepip way of installing a recent version of the pip package manager. You have to install the python3-pip and python3-venv packages.

sudo apt install python3-pip python3-venv

pipx: command not found

python3 -m pipx ensurepath is only able to fix the PATH environment for some (mostly bourne-like) shells. If you are using bash/sh/fish/zsh it may be sufficient to restart your terminal to pick up the new path.

With csh/tcsh you will probably have to add the following to your .login or .cshrc files and/or run rehash to pick up any new binaries.

set path = ( $path $HOME/.local/bin )

ERROR: Could not find a version that satisfies the requirement olive2022 (from versions: none)

Because Olive2022 depends on Python-3.7 or newer, installation fails with this error when the default Python interpreter is older. On Ubuntu 18.04 you can install a newer Python interpreter and explicitly specify it as an alternate interpreter version when installing with pipx.

sudo apt install python3.8
pipx install --python python3.8 olive2022