Common Mistakes

*Common Mistakes in Using Pylauncher and How to Avoid Them

❌ Writing all runs to the same output directory

Problem: Multiple tasks overwrite files or fail unpredictably.

Fix: Ensure every command writes to a unique output folder:

--outDir outCase41
--outDir outCase42

---

❌ Expecting output in tapisjob.out

Problem: You don’t see OpenSees or Python output where you expect it.

Why: PyLauncher captures stdout/stderr per task, not globally.

Fix: Look inside the PyLauncher-created subdirectory in your inputDirectory. Each task has its own log files.

---

❌ Viewing PyLauncher logs in the Data Depot

Problem: Files appear unreadable or empty.

Why: PyLauncher log files often have no file extension.

Fix: Use JupyterHub to inspect PyLauncher output files.

---

❌ Forgetting to disable MPI

Problem: Jobs hang, crash, or behave inconsistently.

Fix: Set:

UseMPI = false

when submitting the job.

---

❌ Using PyLauncher for tightly coupled workflows

Problem: Jobs fail or produce incorrect results.

Why: PyLauncher is for independent serial tasks, not coupled simulations.

Fix: Use MPI directly (outside PyLauncher) for coupled models.