HPC Server: Start Script Quick-Start

HPC Server: Start Script Quick-Start#

Use this option if you would like more control over the allocation request and the environment in which the jupyter server runs. This option requires that you log into the HPC host to run a custom script. Copy this example and edit to your liking.

Launching in this way will start your custom jupyter server and plumb the port forwarding necessary to make it available to your desktop web browser.

  1. Open a terminal window and log in to your HPC host. If you do not know how to do this, you can consult the HPC User Docs.

  2. Clone the hytest repository into the directory you want to work in.

  3. Request a compute node with a command like salloc -A impd -t 1:00:00 -N 1 srun --pty bash. If you are not in the impd group, you will need to replace impd with a group you are a part of. You may also need to update the number of hours you want to have access to the compute node (in this example, we are requesting 1 hour).

  4. Navigate into the environment_set_up directory (cd environment_set_up/) and execute the jupyter-start.sh script (bash jupyter-start.sh).

  5. The jupyter-start.sh script will direct you on how to start port forwarding (that’s the ssh command that you are directed to run from a new terminal on your desktop). You will want to open a new terminal window on your local computer an run this command. You will need to log in with your AD username and password. The cursor will hang after entering your password. This is expected - you can minimize this window and move on to the next step.

  6. The output of the script will also provide the URL where your PC’s browser will find the jupter server. You can open one of these urls in your browser.

Now you are set up with a Jupyter server connected to the HPC you logged in to.

This option can be useful if you want to build your own conda environment, rather than using the provided HyTEST environment. The Pangeo docs are a good place to start if you would like to set up your own environment using this start script. The central HyTEST environment is built using this environment definition. We recommend that you use it as the baseline if you will be building your own environment.

If you choose to do that, you will want to edit the jupyter-start.sh script to reflect the particulars of your custom conda environment.