docker.rst

Building Drake in a Docker Container

Note

These instructions and the Drake Docker images are provided as a courtesy for developers and users and are not covered by continuous integration. Support is on a best-effort basis.

However, there are alternative downstream usages of Drake within a Docker container:

• MIT 6.832 (Underactuated Robotics) Spring 2019 Drake Docker Instructions
• MIT 6.881 (Intelligent Robot Manipulation) Fall 2018 Drake Docker Instructions
• Spartan's Docker Build

Installing Docker and Building and Running a Drake Docker Image

Ubuntu 18.04 (Bionic Beaver)

Install Docker from the Ubuntu package archive:

sudo apt-get update
sudo apt-get install --no-install-recommends docker.io
sudo usermod --append --group docker ${USER}

Reboot or log out and log back in again. Clone the Drake Git repository and build and run a Drake Docker image:

sudo apt-get install --no-install-recommends git
git clone https://github.com/RobotLocomotion/drake.git
cd drake
docker build --file setup/ubuntu/docker/bionic/Dockerfile --tag drake .
docker run --interactive --tty drake bash

Ubuntu 18.04 (Bionic Beaver) with Proprietary NVIDIA Driver 390 and CUDA 9.1 Support

Warning

You MUST first uninstall all previous versions of the proprietary NVIDIA driver, CUDA, and the NVIDIA container runtime for Docker. FAILURE TO DO SO MAY LEAVE YOUR SYSTEM IN A BROKEN STATE. Please carefully read all instructions and proceed with caution. If in doubt, consider following the instructions in the :ref:`previous section <ubuntu_1804_bionic_beaver>` to install and run Docker without proprietary NVIDIA driver and CUDA support.

Install Docker CE from the Docker package archive:

sudo apt-get update
sudo apt-get install --no-install-recommends \
  apt-transport-https \
  ca-certificates \
  gnupg \
  passwd
sudo apt-key adv --fetch-keys https://download.docker.com/linux/ubuntu/gpg
echo 'deb [arch=amd64] https://download.docker.com/linux/ubuntu bionic stable' \
  | sudo tee /etc/apt/sources.list.d/docker.list
sudo apt-get update
sudo apt-get install --no-install-recommends docker-ce
sudo usermod --append --group docker ${USER}

Refer to the warning at the beginning of this section and uninstall all previous versions of the proprietary NVIDIA driver, CUDA, and the NVIDIA container runtime for Docker.

Install the proprietary NVIDIA driver 390 and CUDA toolkit 9.1 from the Ubuntu package archive:

sudo apt-get install --no-install-recommends \
  nvidia-cuda-toolkit \
  nvidia-driver-390

Install the NVIDIA container runtime 2.0 for Docker from the NVIDIA package archive:

sudo apt-get install wget
sudo apt-key adv --fetch-keys https://nvidia.github.io/nvidia-docker/gpgkey
wget -O - https://nvidia.github.io/nvidia-docker/ubuntu18.04/nvidia-docker.list \
  | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update
sudo apt-get install --no-install-recommends nvidia-docker2
sudo pkill -SIGHUP dockerd

Reboot or log out and log back in again. Clone the Drake Git repository and build and run a Drake Docker image:

sudo apt-get install --no-install-recommends git
git clone https://github.com/RobotLocomotion/drake.git
cd drake
docker build \
  --file setup/ubuntu/docker/bionic/Dockerfile.nvidia-cuda-9.1-devel-ubuntu18.04 \
  --tag drake .
docker run --interactive --runtime=nvidia --tty drake bash

Other Platforms

Follow the instructions on the Docker website to install stable Docker Community Edition.

Clone the Drake Git repository and build and run a Drake Docker image:

git clone https://github.com/RobotLocomotion/drake.git
cd drake
docker build --file setup/ubuntu/docker/bionic/Dockerfile --tag drake .
docker run --interactive --tty drake bash

Running a Drake Docker Image with Graphical Interface Support

xhost +local:root; docker run \
  --env=DISPLAY \
  --env=QT_X11_NO_MITSHM=1 \
  --interactive \
  --ipc=host \
  --privileged \
  --tty \
  --volume=/tmp/.X11-unix:/tmp/.X11-unix:rw \
  drake; xhost -local:root

xhost +local:root and xhost -local:root allow and remove access for non-network connections to your local X server and pass the necessary X11 parameters for the graphical display of programs within the Docker container.

Use --runtime=nvidia if you built your image with proprietary NVIDIA driver and CUDA support:

xhost +local:root; docker run \
  --env=DISPLAY \
  --env=QT_X11_NO_MITSHM=1 \
  --interactive \
  --ipc=host \
  --privileged \
  --runtime=nvidia \
  --tty \
  --volume=/tmp/.X11-unix:/tmp/.X11-unix:rw \
  drake; xhost -local:root

The default command will start drake-visualizer and run the bowling ball simulation.

Useful Docker Documentation

• docker build command reference (building an image from a Dockerfile)
• docker cp command reference (copying files and/or folders between a container and the local filesystem)
• docker run command reference (running a command in a new container)
• docker volumes guide (persisting data generated by and/or used by Docker containers)