Setting up a Landlab kernel on the Blanca JupyterHub

[mdpiper]

How can we run notebooks that use Landlab on the Blanca JupyterHub?

We need to:

1. Create a conda environment that includes Landlab
2. Make a Jupyter kernel from that environment

This required a little more work than I expected.

Start by logging in to Blanca and starting a JupyterHub. Then, from the JupyterLab Launcher, start a terminal.

A conda environment that contains Landlab takes up about 5 GB of disk space. I found that I had to modify my conda settings so that my Landlab environment would be installed in my project directory instead of my home directory. I wrote up a brief explanation of how to do this.

Create a conda environment that has Landlab and also ipykernel.

conda create -n landlab -c conda-forge python=3 landlab ipykernel

Once the environment has been created, activate it:

source activate landlab

Then make a kernel:

python -m ipykernel install --user --name landlab --display-name "Landlab"

Now, when you open a new Launcher in JupyterLab, you'll see the Landlab kernel listed. Also, when you open a notebook, you'll see the Landlab kernel listed in the kernel switcher.

[mdpiper]

A short follow-up. If you have a Landlab developer install, you can create a separate Jupyter kernel for it.

Say, for example, that you've followed the instructions for a developer install and placed it in a conda environment named landlab_dev. Start by activating the environment:

source activate landlab_dev

Then install the ipykernel package:

conda install -c conda-forge ipykernel

Then make the kernel:

python -m ipykernel install --user --name landlab_dev --display-name "Landlab Dev"

After a few moments, this kernel should become visible in the JupyterLab Launcher window, and you'll be able to select it when running a notebook.