docker_files

Repository for useful virtualization (docker) stuff. Aimed to make development:

• easy
• simple
• reproducible

How to use repository?

First download and install Docker. Linux is recommended.

Follow instructions using apt-repository.

After docker installation, follow post-installation steps.

How to use hardware acceleration (GPU) with docker?

NVIDIA

Follow instructions for installing nvidia-container-toolkit via /apt.

AMD

Simnilar to NVIDIA, AMD also provides container-toolkit, install it and use it following instructions.

Basic docker usage/steps?

1. Build Dockerfile into docker image using docker build command.
2. Use docker image to run docker container (use first_run.sh script for recommended init configuration)
3. Use docker container for development (use docker start for starting container, and docker exec to enter running container)
4. After done with development use docker stop to stop docker container

Building docker image

docker build -t <img_name>:<tag> 

Running docker container (first time)

Each Dockerfile has corresponding first_run.sh script that can be used. Each first_run.sh contains following command:

#!/bin/bash

CONTAINER_NAME=arm_api2_cont_test
IMAGE_NAME=arm_api2_img:humble

# Hook to the current SSH_AUTH_LOCK - since it changes
# https://www.talkingquickly.co.uk/2021/01/tmux-ssh-agent-forwarding-vs-code/
ln -sf $SSH_AUTH_SOCK ~/.ssh/ssh_auth_sock

docker run \
  -it \
  --network host \
  --privileged \
  --gpus all \
  --volume /dev:/dev \
  --volume /tmp/.x11-unix:/tmp/.x11-unix \
  --volume ~/.ssh/ssh_auth_sock:/ssh-agent \
  --env SSH_AUTH_SOCK=/ssh-agent \
  --env DISPLAY=$DISPLAY \
  --env TERM=xterm-256color \
  --name $CONTAINER_NAME \
  $IMAGE_NAME \
  /bin/bash

Some remarks:

• --network host is used to enable docker container to access local PC network, i.e. see all ROS topics/services that local machine sees (from security perspective, potenital vulnerability)
• --privileged container has sudo privileges (i.e. can access all devices for example, cameras etc...)
• --volume /dev:/dev we map all local devices inside the container (to see cameras and sensors)
• --gpus all docker container has access to all GPUs that PC has (for ML stuff)
• --volume /tmp/.x11-unix:/tmp/.x11-unix remap X11 server for the GUI applications

Starting container

After build --> run we can use start command to start existing container on our PC.

Use docker start -i <cont_name> to start container.

Executing container

If you want to access container that is already started in different bash (terminal), you can use:

docker exec -it <cont_name> bash

Stopping container

You can stop running container with:

docker stop <container_name> 

Important remark

If you stop container, everything you've changed will remain in your container. If you delete container, all progress will be lost. Therefore, delete container only if you've backed up your changes using GIT or something else.

In-depth Instructions

Instructions on how to use following repository can be found here.