If you like to edit code in Jupyter Notebook, the best strategy is using it on your laptop and transferring your code to the cluster when it is ready.
On the cluster, use qsub to run your code.
- On your laptop, make sure you are ssh’d(logged in) into the cluster
- Next startup JupyterNotebook on your laptop and inside JupyterNotebook and navigate to c009.colfaxresearch.com
- JupyterNotebook will connect to the cluster computer node
Alternatively, if you want to store the credentials in your browser, you can log in with a username and a password:
- In JupyterNotebook, navigate to c009.colfaxresearch.com
- Username: u3x09(example)
- Password: 02f51125-cea6-4c73-8f03-118a02b06888(example)
The Jupyter Notebook will open on one of the available compute nodes.
If you are getting a message about connection failure, all Jupyter Hub slots are currently in use. The duration of your Jupyter Notebook session is 1 hour. After that, the server will close, but your files will be autosaved (however, to be safe, save your work periodically).
We do not recommend this method of connection to the cluster because you get only one compute node. As a consequence, you miss out on the opportunity to run multiple jobs at once and experience a parallel speedup.
Do not run Jupyter Notebook servers on the login node. There is very little performance in the login node, and, in addition, your jobs will periodically get killed when they exceed the memory or CPU time quota.
If you must have a Jupyter Notebook, we support a JupyterHub service. To log in, click the button "Start a Jupyter Notebook" found in your Welcome email, under the Learn section.
This document was generated from the following discussion: How to connect to JupyterNotebook on DevCloud