JupyterHub at NCAR

Getting started | Python environments and kernels | Related documentation

The JupyterHub deployment that CISL manages allows "push-button" access to NCAR's Cheyenne supercomputing resource and the Casper cluster of nodes used for data analysis and visualization, machine learning, and deep learning.

It gives users the ability to create, save, and share Jupyter Notebooks through the JupyterLab interface and to run interactive, web-based analysis, visualization and compute jobs on Cheyenne and Casper. JupyterHub is an alternative to X11 access for interacting with those resources to run jobs as well as for using web-based interactive shell functionality without the need to install or use software such as SSH or PuTTY.

This JupyterHub deployment is a test system that is subject to more planned and unplanned downtimes than other resources in the NCAR/CISL high-performance computing ecosystem, such as login nodes. Responding to outages and providing updates to JupyterHub may be delayed while CISL staff support our full production HPC systems.
While JupyterHub provides added value and convenience, users are advised to be familiar with our Jupyter and IPython documentation as an alternative for continuing their work with HPC resources when JupyterHub is not available.


Getting started

Use your web browser to go to jupyterhub.ucar.edu. Chrome and Firefox are recommended for all users.

Select one of the available NCAR resources: the Cheyenne supercomputer or the Casper cluster.

Buttons to select either Cheyenne or Casper cluster

Log in with your NCAR username and Duo two-factor authentication, just as you would when logging directly in to either system.

JupyterHub login panel

Launch the server if it is not running.

Launch Server panel

Use the form provided (images below) to name your job, specify your project code, and adjust job options as needed. For more information about the options, see:

Launch your job when ready. This job only gives you access to the JupyterLab instance. If you need more resources, you can launch another job or jobs from within JupyterLab.

Form for starting Cheyenne job Form for starting Casperjob

After launching the job, you will have access to multiple kernels in the web interface (image below) for working with various languages and applications.

Note that the “File browser” icon (upper-left of following image) allows you to explore your home directory only. To change to your scratch or work space, create soft links in your home directory to those locations. 

view of JupyterHub web interface


Python environments and kernels

The JupyterLab dashboard provides access to Notebook and Console kernels, which are programming language interpreters. Available kernels, which change periodically as new releases are installed, include:

  • Multiple Python 3 interpreters with varying package support including a basic install (Python 3), the Pangeo stack installed with conda (Pangeo), and the NCAR Package Library (NPL) that is otherwise available via environment modules
  • R
  • Julia
  • Three C++ interpreters with different standards compliance
  • A Bash interpreter that provides a shell environment

Related documentation

See these related CISL documentation pages for additional support:

Python - NCAR Package Library

This applies to using Python and the NPL virtual environments via this JupyterHub deployment as well as by logging in directly to Cheyenne or Casper. This section of the documentation describes how to clone an environment and create Jupyter kernels.

Jupyter and IPython

See this section for information about using Python kernels with QtConsole.

Conda environments on JupyterHub

Describes how to activate Conda environments that you create so you can access them through the NCAR JupyterHub.