Running Jupyter Lab from within Docker

This runs a Jupyter lab within Docker and listens on port :5150 on the Docker host. The image builds upon the existing Docker image of Jupyter Notebook by adding in kernels for Julia and Bash, among others. It also saves any created Jupyter notebooks on a shared folder with the host. This allows for notebooks to persist across different Docker container instances.

See https://hub.docker.com/repository/docker/rwcitek/jupyter-notebook on DockerHub.

Pull image ( optional )

docker image pull rwcitek/jupyter-notebook:latest

Launch container

DooD=""
#  Uncomment next line to enable the socket for Docker-out-of-Docker
#  DooD="-v /var/run/docker.sock:/var/run/docker.sock"

SHARED=~/shared.jupyter-lab
mkdir -p "${SHARED}"
docker \
    run \
    -d \
    -p :5150:8888 \
    -e JUPYTER_ENABLE_LAB=yes \
    -v "${SHARED}":/home/jovyan/shared ${DooD} \
    -w /home/jovyan/shared \
    --name jupyter-lab \
    rwcitek/jupyter-notebook:latest

host=192.168.1.8         # On the Mac ( the IP of any interface on the host )
host=127.0.0.1           # On a Chromebook or remote cloud instance using ssh tunneling ( -L :5150:127.0.0.1:5150 )

while true; do
  token=$( docker container logs --since 5s jupyter-lab 2>&1 | grep -m1 -o token=.* )
  [ "${token}" ] && echo -e "\n\n\nhttp://${host}:5150/lab?${token}\n\n\n" && break
  sleep 2
done |
tee /tmp/jupyter.url.token.txt

Customize

docker exec -i jupyter-lab /bin/bash -c 'cat > ~/.bash_aliases' <<'eof'
  alias cls='clear';
  alias dir='ls -la';
  alias h='history';
  alias more='less -iX';
  export HISTCONTROL=ignoredups:ignorespace;
  export HISTFILESIZE=50000;
  export HISTSIZE=50000;
  export HISTTIMEFORMAT='%t%F %T%t';
  export PAGER='less -iX ';
  export IGNOREEOF=20;
  export PS1='\u@\h: \w\n\$ ';
eof

Update system packages

Eventually, when the packages become too far out of date, I will update the base image. But for now, this is a way to update the instance.

<<'eof' docker exec -i -u root jupyter-lab /bin/bash
    export DEBIAN_FRONTEND=noninteractive
    apt-get update
    apt-get dist-upgrade -y
    apt-get install -y --no-install-recommends \
        csvkit \
        ;
eof

Enabling Docker-out-of-Docker

To enable "Docker-out-of-Docker", you will need to enable access to the socket ( see above comment during launch ). Docker is already installed. See the Dockerfile or this DooD gist for installation details.

DooD allows you to manage Docker resources on the host from within a container instance. This definitely has some security risks. But if you own the host and are the only user on the host, it can be extremely useful.

Update pip modules

Eventually, when the modules become too far out of date, I will update the base image. But for now, this is a way to update the instance.

<<'eof' docker exec -i jupyter-lab /bin/bash
    # you'd think there'd be a cleaner way to upgrade all installed packages
    pip install --upgrade $( pip list | awk 'NR > 2 {print $1}' )
    pip install \
      dtale \
      mitoinstaller \
      openpyxl \
      wquantiles \
    ;
    python -m mitoinstaller install
eof

Blog post about D-Tale

Blog post about mito

Container operations

Pausing

Pausing the container temporarily "inactivates" it. It no longer responds to requests. However, as a process, it is still considered running. The upshot is that the up-time is unaffected and the same URL will work for connecting to it once you unpause it.

docker container pause jupyter-lab
docker container unpause jupyter-lab

# To redisplay the existing URL
host=192.168.1.8         # On the Mac ( the IP of any interface on the host )
host=127.0.0.1           # On a Chromebook or remote cloud instance using ssh tunneling ( -L :5150:127.0.0.1:5150 )

token=$( docker container logs jupyter-lab 2>&1 | tac | grep -m1 -o token=.* )
echo -e "\n\n\nhttp://${host}:5150/lab?${token}\n\n\n"

Stopping

Stopping the container actually shuts it down. As a process, it is no longer running and therefore no longer responds to requests even though the filesystem for the container still exists. The upshot is that upon restarting, a new URL is created.

docker container stop jupyter-lab
docker container start jupyter-lab

# Wait for the new URL
host=192.168.1.8         # On the Mac ( the IP of any interface on the host )
host=127.0.0.1           # On a Chromebook or remote cloud instance using ssh tunneling ( -L :5150:127.0.0.1:5150 )

while true; do
  token=$( docker container logs --since 5s jupyter-lab 2>&1 | grep -m1 -o token=.* )
  [ "${token}" ] && echo -e "\n\n\nhttp://${host}:5150/lab?${token}\n\n\n" && break
  sleep 2
done |
tee /tmp/jupyter.url.token.txt

Commiting

Commiting the container creates a new image from the existing container. This is not recommended, but can be useful in a pinch. The container can be in any state: running, paused, or stopped. The commit will pause the container and then unpause it upon completion. In this example, the epoch time is used as the image tag.

docker container commit jupyter-lab jupyter-lab-commit:$( date +%s )

Removing

Removing the container does just that. The container has to be stopped first.

docker container stop jupyter-lab
docker container rm jupyter-lab

Dockerfile

Since the Dockerfile is included with the image, it can be copied out of the image.

docker \
    run \
    --rm \
    --entrypoint cat \
    rwcitek/jupyter-notebook:latest \
    /Dockerfile |
  tee Dockerfile

Docker build and push to Dockerhub

Once you have the Dockerfile, you can modify it and build a custom image.

tag=$( date +%s )
docker build --tag rwcitek/jupyter-notebook:${tag} .
docker image tag rwcitek/jupyter-notebook:${tag} rwcitek/jupyter-notebook:latest
docker login
docker push rwcitek/jupyter-notebook:${tag}
docker push rwcitek/jupyter-notebook:latest

Or build from the Dockerfile in Github

curl -s https://raw.githubusercontent.com/rwcitek/Docker.jupyter-lab/main/docker/Dockerfile |
docker image build --tag rwcitek/jupyter-notebook:latest -