Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.


The jup_sched utility has been deprecated. Please refer to Getting started with HAL OnDemand for instructions to use Jupyter Notebook with HAL OnDemand

titleDo not run Jupyter directly on the login node

You will not get GPUs this way and will also impact others' ability to use the system. Refer to the link above

Table of Contents

Code Block
$ jup_sched -h
usage: jup_sched [-h] [-e EMAIL] [-t TIME] [--gpu] [-py2] [--password] [-d]

Jupyter Notebook Scheduler

optional arguments:
  -h, --help            show this help message and exit
  -e EMAIL, --email EMAIL
                        specify an email address for the scheduler
  -t TIME, --time TIME  specify the running time for the scheduler in the
                        format HH:MM:SS
  --gpu                 use this option if you require a GPU to be scheduled
  -py2, --python2       use a Python 2.x environment (Python 3.x is used by
  --password            re-print the password and address for the jupyter
                        notebook WILL IGNORE ALL OTHER ARGUMENTS
  -d, --debug           keep the scheduler's transcript of the console output
                        (off by default)



To get a simple Jupyter Notebook running on one CPU core on a compute node:

Code Block
$ jup_sched
[jupyter_scheduler INFO] no time on arguments, scheduling notebook for 1 HOUR
[jupyter_scheduler INFO] Waiting for scheduler...

Please direct your browser to:
Use the password to login:

Run "jup_sched --password" to view this message again

Open a browser on your local computer (not HAL), navigate to the provided address and use the provided password to log in. The job will terminate when the time limit is reached or if you exit via the web interface.

To get a Jupyter Notebook for two hours:

Code Block
jup_sched -t 02:00:00

To get a Jupyter Notebook with a GPU:

Code Block
jup_sched --gpu