Jupyter

SCRP relies on Jupyter and JupyterHub to provide browser-based access.

Basics

Navigate to one of the following URLs on a browser:

• https://scrp-login.econ.cuhk.edu.hk/jupyter
• https://scrp-login-2.econ.cuhk.edu.hk/jupyter

This launches a Jupyter instance on the corresponding login node. Below is a list of software we support on Jupyter:

Software	Graphical User Interface	Jupyter Notebook	Console
Julia	-	Yes	Yes
MATLAB	Yes	Yes	Yes
Python	-	Yes	Yes
R	Yes	Yes	Yes
Stata	No	Yes	Yes

Running Jupyter on a Compute Node

To launch Jupyter on a compute node, simply navigate to one of the following URLs:

• https://scrp-login.econ.cuhk.edu.hk/slurm
• https://scrp-login-2.econ.cuhk.edu.hk/slurm

Once you log in, you will see a list of predefined options with different resource configurations.

Important: when you have completed your work, please shut down the server by going to [File] > [Hub Control Panel] > [Stop My Server]. Without taking this step, your Jupyter server will keep running and count towards your resource quota.

As this option is primarily intended for teaching and code testing, we are currently setting a hard-coded limit of three hours per session. You can access the full capability of the cluster through a terminal as before.

Launching a Private Jupyter Instance through SSH

Private Jupyter Instance on a Login Node

Steps to run your own Jupyter instance on a login node through SSH, using scrp-login as an example:

1. Launch jupyter with private-jupyter.
2. You should see a URL http://localhost:xxxx/?token=... print on screen.
3. Set up SSH port forwarding:

   • On Linux, Mac OS or Windows Subsystem for Linux, connect to the login node in the following way:

     ssh -L xxxx:localhost:xxxx username@scrp-login.econ.cuhk.edu.hk
     

     where xxxx is the four digit number after localhost: in step 2.

   • If you use PuTTY follow instructions here.

4. On a browser, navigate to the URL you see in step 2.

Private Jupyter Instance on a Compute Node

Steps to run your own Jupyter instance on a compute node through SSH:

1. Launch jupyter with compute private-jupyter. Adjust Slurm options as needed.
2. You will need two pieces of information:
   • The name of the compute node you are connected to. This is the first line print on screen.
   • The URL http://localhost:xxxx/?token=....

3. Set up SSH port forwarding. On Linux, Mac OS or Windows Subsystem for Linux, you can set it up in the terminal:

   ssh -t -L xxxx:localhost:xxxx username@scrp-login.econ.cuhk.edu.hk ssh -L xxxx:localhost:xxxx compute-node-name
   

   where xxxx is the four digit number after localhost: and compute-node-name is what you saw on the first line in step 2.

4. On a browser, navigate to the URL you see in step 2.