Note: This documents the legacy per-user PBS workflow. For the recommended shared server approach, see Shared server startup — just run bash hpc_utils/connect.sh.

Follow the instructions below to submit a job on Randi to start the visualization app and terminate it after use.

Preliminary steps¶

Copy hpc_utils/start_app_on_randi.pbs to your user folder.
Update the output and error file locations (lines 3 and 4) to point to your own log directory.
Update the conda environment path on line 18 if your environment is in a different location.
Update the project directory path on line 20 to match your codebase location.

Starting and viewing the app¶

We use a SLURM job to run the app on a compute node.

To start the app, go to the folder where you copied start_app_on_randi.pbs and execute:

sbatch start_app_on_randi.pbs

Get information on the node your job is running on (the node will be shown in the NODELIST column):

squeue -u yourusername

Open a new terminal and create an SSH tunnel to the compute node on port 5601:

ssh -N -f -L 5601:nodename:5601 yourusername@randi.cri.uchicago.edu

Open an internet browser on your local computer and go to: http://localhost:5601/app

You will be prompted to log in. Use the credentials defined in credentials.json in the project root.

To stop the app, find the SLURM job id and cancel it:

squeue -u yourusername
scancel jobid

Troubleshooting SSH tunnelling¶

If the port is unavailable for tunneling, find and kill the process holding it:

ps aux | grep 5601

Get the process id from the output and kill it:

kill -9 processid