When working with Jetpack 4.5.1 for NVIDIA Jetson Nano, it comes with Docker installed, but not docker-compose
. And because the Nano is arm64 CPU architecture, installing Docker Compose is not straight forward. There are a number of existing tutorials listed in the references section, but none of them worked exactly as-is for me, or some of them have slightly bloated or less secure instructions that I would prefer. So here is my lean, mean, securely-installing machine instructions.
Install prerequisites
Make sure your Ubuntu installation is up to date with the usual apt-get commands
sudo apt-get update && sudo apt-get upgrade -y
Install necessary Ubuntu packages
sudo apt-get install -y python3-pip libssl-dev python-openssl libffi-dev rustc cargo
Now the default version of pip and setuptools that ships with Jetpack 4.5.1 is not new enough to successfully install docker-compose
. Those need to be upgraded before continuing. You can upgrade to the newest version, but here are the versions that I tested.
python3 -m pip install --upgrade pip==21.1.3 setuptools==56.2.0
Then, once those are upgraded you can install Docker Compose. Docker Compose needs to be installed as root to give it access to the low-level system resources it needs to execute the installer.
sudo -H python3 -m pip install docker-compose==1.29.2
And after that completes, you should be able to get the version with docker-compose --version
Docker without sudo
One thing that can be really annoying when using the default Jetpack version of docker is that you have to run everything with sudo
. Here are the steps to add your user as part of the docker group.
sudo groupadd docker
sudo usermod -aG docker $USER
Then log out and log back in as your user to make sure all the settings are applied properly. No more sudo!
References
https://developer.nvidia.com/jetpack-sdk-451-archive
https://docs.docker.com/engine/install/linux-postinstall/
https://collabnix.com/running-docker-compose-on-nvidia-jetson-nano-in-5-minutes/
Top comments (0)