videodrone 0.9.5.post1

WebRTC python unit tests inspired to '80s

VideoDrone

How many connected users your favourite open source video conferencing solution can handle? That's the question that videodrone wants to answer. Empirically.

Videodrone tests popular WebRTC Platforms with Selenium HQ and Python, aiming to be a lightweight build system for load test orchestration.

At this moment only chrome driver was implemented and well tested on Debian based distributions, fill free to contribute with your "drone connector". See Drone Connectors for further informations.

Requirements:

• python {36,37,38}
• selenium
• chromium web browser

Available drones:

• BBB meet GARR, bbbmeet_garr_chrome
• Pexip GARR, pexip_garr_chrome
• jitsi, jitsi_chrome
• Edu meet GARR, edumeet_garr_chrome
• Open meet GARR, open_garr_chrome
• lepida #iorestoacasa work, lepida_iorestoacasawork_chrome

You should know, that:

• WebRTC needs a huge load, on a intel i7 processor you would start from 5 to 8 drones
• hundreds of drones should be handled in a distribuited environment, a cluster of videodrones

Setup

Provide a fully working python3 pip environment, with virtualenv installed in. You can even use build.sh to build your videodrone project.

Prepare environment

apt install python3-pip wget chromium unzip
pip3 install --upgrade pip
pip3 install virtualenv

wget https://raw.githubusercontent.com/peppelinux/videodrone/master/build.sh -O build.sh
bash build.sh VideoDrone

You can even install videodrone by hands.

mkdir VideoDrones && cd VideoDrones
virtualenv -ppython3 env && source env/bin/activate
pip install videodrone

Create the following directories before executing videodrone.

• y4m, where your preferred y4m files resides. They will be used randomically.
  • mkdir y4ms
  • wget https://media.xiph.org/video/derf/y4m/students_cif.y4m -O y4ms/students_cif.y4m
• drivers, where your selenium drivers resides.

Setup in LXC container

apt install lxc
CONTAINER_NAME=deb10
lxc-create -t download -n $CONTAINER_NAME -- -d debian -r buster -a armhf
lxc-start deb10
lxc-attach deb10

# then choose your preferred setup as show in the previous sections.

Docker Image

docker image build --tag videodrone .

# go in
# docker container run -it videodrone /bin/bash

# run the container with your preferred configuration
docker container run -dit -e VIDEODRONE_DRIVER=/usr/bin/chromedriver videodrone videodrone -room thatroom -c videodrone.drones.jitsi_chrome -y4m /VideoDrone/y4ms/ -lifetime 33 -n 4

Run

VIDEODRONE_DRIVER environment variable can override the default path, which is $VDPATH/drivers/chromedriver.

example, this connector is configured to create a single drone to "https://meet.jit.si/thatroom":

VIDEODRONE_DRIVER=/usr/bin/chromedriver videodrone -c "videodrone.drones.jitsi_chrome" -r thatroom -y4m ./y4ms/

There will be a party of 4 drones in "thatroom":

videodrone -room thatroom -c "videodrone.drones.edumeet_garr_chrome" -y4m ./y4ms/ -n 4

Output

INFO:__name__:Started drone <Process name='Process-1' pid=25900 parent=25898 started>
INFO:__name__:Started drone <Process name='Process-2' pid=25960 parent=25898 started>
INFO:__name__:Started drone <Process name='Process-3' pid=26155 parent=25898 started>
INFO:__name__:Started drone <Process name='Process-4' pid=26373 parent=25898 started>
INFO:__name__:Drone destroyed
INFO:__name__:Drone destroyed
INFO:__name__:Drone destroyed
INFO:__name__:Drone destroyed

What happens

Another example with Garr BBB meet. headless set to false means that the browser will run in foreground.

./videodrone -c "videodrone.drones.bbbmeet_garr_chrome" -room "/b/roomname -pin thatpin -y4m y4ms/ -headless 0 -lifetime 27

Drone Connectors

Drone connectors are selenium browser macros, written in python, as simple as possibile. Drone connectors must be packaged and installed or included in your PYTHONPATH (sys.path). You can even create a python local package folder, in your VideoDrone Project.

See videodrone.drones.jitsi_chrome for example.

Credits

Fabio Farina (Garr Consortium), Massimo Carboni (Garr Consortium), Garrlab community.