Creating Jupyterhub on Azure using a Jupyter Notebook

This is a short tutorial on installing Jupyterhub using a Jupyter notebook. You must have Azure the Azure CLI, Helm, and Conda installed locally.

See instructions for installing the Azure CLI here. See instructions for installing the Helm Client here. See instructions for installing the Conda here.

We will start by installing carme.


This will activate the virtual environment with Carme installed. If in the future you log in again this will not be installed but not active. To activate the environment, just use the commands below:

source activate carme-env

To verify that Carme is installed and working type carme --help. This will give an overview of the carme package.

Usage: carme [OPTIONS] COMMAND [ARGS]…

Options: –version Show the version and exit.

–help Show this message and exit.

Commands: build Build project docker images. cleanup Delete and remove all images. cmd Runs commands from the commands folder. convert Launch Jupyter Notebook (using Docker). git Initializes git and performs initial… lab Launch Jupyter Notebook (using Docker). new Creates a new carme project in project_dir or… notebook Launch Jupyter Notebook (using Docker). package Manage packages on the project. save A simler alias to git commit and push. start Launch all installed applications. stop Stop (and optionally remove, –remove) all…

Create Your First Carme Project

The next thing that we are going to do is to create a new Carme project based on the az-z2jh package. The az-z2jh project includes the additional packages for creating a Kubernetes cluster (az-cluster), installing JupyterHub (jupyterhub), and customizing the singleuser container (‘jupyter-cpu’). You can find the documentation for the az-z2jh project here. The az-z2jh package is based on the Zero to JupyterHub project, which has extensive documentation on launching JupyterHub on a variety of cloud platforms. Carme just makes it a bit easier.

This will create the carme project:

carme new mycluster --package az-z2jh

Now change to the Carme project directory and look at the directory structure:

cd mycluster

From here launch the Jupyter using:

jupyter lab

Or jupyter notebook if you prefer.

The notebook for launching Jupyterhub is under: code/notebooks/az-cluster.ipynb

GO THERE NOW! The rest of this tutorial is in the notebook. Just press play to continue down the process of installing.

More tutorials coming soon. Join the Slack channel to stay up to date!

Back to blog

Containers increase reproducablity and scalablity, and Carme helps data scientists to use containers.

Working with containers allows your infrastructure to become code [this is often referred to as or infrastructure as code (IAC)]. This ensures that others will be able to reproduce your work, and the Carme package system makes sharing your work a snap. Containers are also useful when you want to scale an analysis, perhaps moving from your local machine to a GPU in the cloud or from a single GPU server to a Kubernetes cluster for a classroom Jupyterhub instance.