Kubernetes Job and CronJob

A Job manages finite work toward successful completion. A CronJob creates Jobs according to a schedule. Their controller state describes execution management, not an atomic transaction with an external data destination.

CronJob concurrencyPolicy Forbid applies to Jobs from that same CronJob. It does not lock a table against another CronJob or a manually launched Job. A late schedule can be skipped according to the starting deadline.

Programs can execute more than once, so keep input identity and repeat-safe destination effects explicit. Job retry limits and execution deadlines control attempts; they do not roll back an already published result. History cleanup is separate from business-data retention.

For a non-parallel Job, restartPolicy Never means a failed container is not restarted inside the same Pod; the Job controller can create another Pod subject to its retry policy. backoffLimit bounds retries, while activeDeadlineSeconds limits active Job duration. A CronJob’s startingDeadlineSeconds instead decides whether a late scheduled start is still eligible. These clocks answer different questions.

See Running Data Workloads on Kubernetes for manifest examples.

Further reading

• Kubernetes: Jobs
• Kubernetes: CronJob