RViz on docker container

This example shows how to run RViz in a Docker container.d

Quick start

To run example, type the following commands:

cd examples/ros2/nvidia-container-toolkit
docker compose build
docker compose up -d

This example shows how to run GUI app like RViz in a Docker container. There are several ways to run RViz in a container in linux systems:

Run container without --net=host option, this is the most secure way to run RViz in a container, but it requires to mount /tmp/.X11-unix to the container.
Run container with --net=host option, in this case the network is shared between the host and the container. This is the easiest way to run RViz in a container, but it is not secure.

On both cases, you will need to set the following environment variables inside the container before run GUI apps:

You can use the NVIDIA Container Toolkit to run RViz in a container with GPU support. To do so, you will need follow the steps below:

Select runtime to nvidia in docker-compose.yml file.
Set the follwing environment variables inside the container at startup:
- NVIDIA_VISIBLE_DEVICES: to one or more comma-separated GPU UUIDs, or all to use all GPUs. Also you can use numbers to select GPUs by index.
- NVIDIA_DRIVER_CAPABILITIES: to one or more comma-separated driver capabilities. For RViz, you will need to set graphics and utility. To use all capabilities, set it to all.

To get the GPU UUIDs, you can run the following command:

nvidia-smi --query-gpu=uuid --format=csv,noheader

The easiest way to run GUI app in a container is to use the --net=host option without NVIDIA support. See docker-compose.yaml for more details.