The batch job that never finished
A scheduled job runs every fifteen minutes. Its own work takes about forty seconds, and a log-shipping helper runs beside it in the same pod so the output reaches the log store.
The pod never reaches Completed. The job's container exited on time; the helper is still running, because a log shipper has no natural end. The schedule refuses to start a run while the previous one is still going, so every run after the first has been skipped.
Boundaries, outermost first: Pod (one run): Job container (exits after 40s), Log shipper (has no natural end; FAILED: still running), an empty slot for the how the helper is declared Outside every boundary: Log store, Schedule (every 15 minutes; FAILED: runs skipped) Connections: Schedule controls Job container — creates the run (step 1) how the helper is declared controls Log shipper (step 2) Log shipper publishes to Log store — ships output