my docker cheatsheet

docker_cheatsheet.md

basics

• check that docker is working: docker run hello-world
• prune images: docker image prune -a
• prune containers: docker container prune
• check nvidia-docker2 is working (with CUDA 10.0 in base): docker run --rm --gpus all nvidia/cuda:10.0-runtime nvidia-smi
• build with Dockerfile: docker build -t your_image_name .
• run built image as container: docker run --name your_container_name your_image_name (for extra options see here)
• kill running container: docker kill your_container_name

Mount a volume into the docker container

use the -v flag, You need to give the absolute local path or a volume name and map it to a directory within the container -v <source>:<target>. See this stackoverflow post or this article on freecodecamp for more details.

docker run -it --rm -v $(pwd)/dir/to/share:/container/path/to/shared/dir --name your_container_name your_container_id

and instead of mounting your local directory, we can use docker volume, which is fully managed by docker... see this guide for more details

Make Docker container start automatically on system boot

docker run --restart=always

Note that with this you cannot use the option --rm. Credit to this stackoverflow post and the official doc

watch a running container

once in interactive mode it's kind of hard to exit without killing the container. The best way to watch a container is actually by combining watch and docker logs. For example:

watch -n 0.1 "docker logs --tail 30 your_container_name_or_id"

Debug something in your Dockerfile

best to just run the base image interactively and try the command in bash:

docker run -it --rm {image:tag} /bin/bash

or to override an image's entry point:

docker run -it --rm --entrypoint /bin/bash {image:tag|image_id}