Jobs and CronJobs

Work that finishes

A job runs pods until a set number succeed, then stops. It is the right object for migrations, imports, and any batch task — unlike a deployment, which will restart your finished task forever.

Cron jobs create jobs on a schedule. Decide explicitly what happens when a run overlaps the previous one; the default allows concurrent runs, which is rarely what a nightly import wants.

Set a retry limit and a deadline. A job that retries indefinitely on a permanent failure quietly burns cluster resources for days.