b. Python and Jupyter notebook

Jupyter is a server-client application that allows editing and running computer code (in our case Python).

On levante.dkrz.de (shell console 1):

• module load python
  jupyter notebook --no-browser

• (You need later the token shown in this shell console)

On your local machine (shell console 2):

• ssh -f levante.dkrz.de -L 1234:localhost:8888 -N

• Then use this URL in in your web browser: https://localhost:1234/ → in the form filed you can paste the token

This section explains how to setup a Jupyter session with the help of JupyterHub, a Jupyter server, that can run notebooks for multiple users.

DKRZ offers documentation about how to use Jupyter on their machines, please take a look at Welcome to Jupyterhub @ DKRZ!.

Use the following steps to setup a jupyter notebook using the Jupyter Hub which is provided by DKRZ:

• Log on to the Jupyter Hub page of DKRZ: https://jupyterhub.dkrz.de/hub/login with your user credentials
• Select either preset options or advanced options.
• For both options you will have to set at least the project account that will be charged for the computing time, usually this account starts with an mh followed by a number.
• Select the slurm queues (if you choose the advanced option). The JupyterHub uses the slurm workload manager to start your Jupyter session. Hence you will have to choose a queue for your computations. All available queues and their hardware specifications can be found on the dkrz user portal. If you choose advanced options you should bare in mind that certain slurm combination settings are not available. For instance jobs running for more than 8 hours on the computing queues are not possible.

After you've clicked the start button and the jupyter server has been setup you'll see the notebook interface. Here you can create a new or load an existing jupyter notebook.