jsiloto / docker-project

General purpose docker boilerplate and cheat-sheet

docker-project

General purpose docker boilerplate and cheat-sheet

Build and Run

Most docker images by default run as root. That may bring unintended consequences as files created within docker will have a root user. To better manage this, this boilerplate provides scripts to build and run your docker container as your own user.

./docker_build.sh
./docker_run.sh -g 0

To change the image edit the Dockerfile file.

Managing your Container

Lets check a container lifetime example

./docker_run.sh -g 0
# Detach from your container by pressing Ctrl+p-q
# Check your container exists
docker ps
# List all images
docker images
# Reattach to your container
docker attach <your container name(use tab)>

Now Open another Terminal

# You can run multiple processes in your container
docker exec -it <your container name(use tab)> /bin/bash
# Kill your process
# Ctrl+C
# Kill your container
docker kill <your container name(use tab)>
# Check your container no longer exists
docker ps

Running with arguments

The last argument in the docker run command is optional and provides you the starting binary of your command application (usually \bin\bash)

Try:

./docker_run.sh -g 0 'jupyter notebook --no-browser -NotebookApp.token='ABC''

Running without ./docker_run.sh

Managing GPUs and CPUs

https://docs.docker.com/config/containers/resource_constraints/

docker run --rm -it --gpus='"device=0,1"' --cpus=2.3 --memory=1GB ${USER}/docker-image

Using the disk

#docker run --rm -it -v local_path:container_path ${USER}/docker-image
docker run --rm -it -v $PWD:/work ${USER}/docker-image

Networking

docker run --rm -it --net=host ${USER}/docker-image \
 jupyter notebook --no-browser -NotebookApp.token='ABC'

docker run --rm -it -p 8123:8888 ${USER}/docker-image \
 jupyter notebook --ip 0.0.0.0 --no-browser -NotebookApp.token='ABC'

DockerHub

You can do the same thing we did building a custom image by exploring existing images in the community. In this case jupyter/datascience-notebook

docker run --rm -it --network host  -v $PWD:/work -w /work jupyter/datascience-notebook jupyter notebook --no-browser -NotebookApp.token='ABC' -ip 0.0.0.0 -NotebookApp.allow_origin='*'

Check out more at https://hub.docker.com/