Example Documentation: Parallel Indexed Job
Goal
Understand parallel batch processing with Indexed Jobs, parallelism and completions.
What Is Done
A Job creates six work units and runs up to three Pods in parallel. Each Pod receives a unique JOB_COMPLETION_INDEX.
Typical Commands
kubectl apply -f manifests/
kubectl wait --for=condition=complete job/indexed-worker -n indexed-job-lab --timeout=120s
kubectl logs -n indexed-job-lab job/indexed-worker --all-containers
Validate Success
- The Job reaches
Complete. - Logs show different shard indexes.
- At most three Pods run at the same time.
Common Mistakes
- Indexed Jobs are confused with normal Job parallelism.
backoffLimitPerIndexis not supported on older clusters.- Workloads do not use the provided index.