Part 1 — Profile & Submit a Job
Yesterday’s script runs, on one filing at a time, on an interactive Yen node you share with everyone else. This part is finding out what it actually costs — and then handing it to the scheduler instead of holding a terminal open.
9:00 to 10:30. A short lecture, then the work block.
First, get back on the Yens. Pick one of the five Yens and open its JupyterHub, then log in with your SUNetID:
yen1 · yen2 · yen3 · yen4 · yen5
Click the blue “+” to open the Launcher and open a Terminal — no
ssh, and no Duo prompt. Blocks labelled JupyterHub Terminal and Yen Terminal both go at that shell prompt; a Jupyter Notebook Cell does not.
Hands-On Lab
| What you’ll learn | Bonus | |
|---|---|---|
| 1. Profile | Measure a script you have never seen, then the real workload over 10 filings | Profile two more scripts · explore real cluster usage data · size up your own machine |
| 2. Document | Put the three numbers in your README, where the next step reads them | Let Claude write it into the README |
| 3. Submit | Read the live queue, then write #SBATCH directives by hand, submit, monitor and cancel |
Other ways to run and inspect jobs · turn this into a Claude skill |
| 4. Read Logs | Submit a job that fails on purpose, and find out why from its .err |
Three more broken jobs · let Claude read the log for you |