Dockerfiles with Jupyter

Hosting jupyter servers in Docker containers can be extremely useful, ~~as this typically doesn’t involve logging into the container at all,~~[1] and you can just connect to your server in a browser.

A typical docker file for this purpose looks like

FROM debian:latest # or whatever base image you want

RUN pip3 install --upgrade pip
RUN pip3 install jupyter jupyterlab #plus whatever else

CMD ["jupyter-lab", "--no-browser", "--allow-root", "--ip=0.0.0.0", "--port=8989"] # The port doesn't have to be the same, just some random port not in use

Running container

Now, with image in hand, navigate to the directory of interest and run the following command:

docker run -d --rm -p <port-ID>:<port-ID> -v $(PWD):/container/dir/path <image-ID>

where the <port-ID> matches the port used in the image (at least the second one) and the “/container/dir/path” directory is whatever you want to call it (directory inside the container). The image (<image-ID>) is obviously the image you want to run (e.g. scottperkins/gwat.jupyter)

Logging in

For now, you have to log into the container with

docker exec -it <container-id> bash

Then, inside the container, run

jupyter server list

Go to localhost:<port-ID> in a browser and copy the token from the docker container to log in.

You’re in! Now you can exit out of the interactive session of the container if you want.