A scheduled or one-shot task that runs to completion, the workhorse of background processing and periodic maintenance.
A job is a unit of work that runs to completion, as opposed to a long-running service that handles requests indefinitely. Cron jobs run on a schedule (every hour, every Monday at 3am); batch jobs run once over a finite input set; Kubernetes Jobs and CronJobs encapsulate both. Common uses: nightly database backups, weekly bill generation, periodic certificate rotation, batch ML inference. Jobs need their own observability story (did it run, did it finish, did it succeed) because they don't generate live request metrics.
Silent job failures are some of the most damaging incidents teams face, the nightly backup didn't run for 30 days, the certificate renewal was supposed to happen quarterly but a bug skipped one, the bill-generation job hung silently for a week. Treat every job as a service with its own SLO (last-success-timestamp, duration, output-size) and alert on a missed run, not just on a failed one.
See the part of the platform that handles job (cron / batch) in production.