Jobs & CronJobs

Run tasks to completion and schedule recurring work.

Time: ~10 minutes Difficulty: Beginner

What You Will Learn

• Jobs: run-to-completion workloads (not long-running services)
• Parallel jobs with completions and parallelism
• Failure handling with backoffLimit and activeDeadlineSeconds
• CronJobs: scheduled, recurring tasks
• Concurrency policies and history limits

Deploy

Navigate to the demo directory:

cd demos/jobs

kubectl apply -f manifests/namespace.yaml

Simple Job

Calculate 2,000 digits of pi and exit:

kubectl apply -f manifests/simple-job.yaml
kubectl get jobs -n jobs-demo -w

When the job shows 1/1 completions, check the result:

kubectl logs job/pi-calculator -n jobs-demo

Parallel Job

Run 5 tasks with 2 workers at a time:

kubectl apply -f manifests/parallel-job.yaml
kubectl get pods -n jobs-demo -l job-name=batch-processor -w

Watch pods run in batches of 2 until all 5 completions are done:

kubectl get jobs batch-processor -n jobs-demo

Failing Job

A job that randomly fails, demonstrating retry behavior:

kubectl apply -f manifests/failing-job.yaml
kubectl get pods -n jobs-demo -l job-name=flaky-job -w

The job retries up to 3 times (backoffLimit: 3) and gives up after 60 seconds (activeDeadlineSeconds: 60). Check which attempts failed:

kubectl get pods -l job-name=flaky-job -n jobs-demo
kubectl logs -l job-name=flaky-job -n jobs-demo --prefix

CronJob

Run a health report every 2 minutes:

kubectl apply -f manifests/cronjob.yaml
kubectl get cronjobs -n jobs-demo

Wait 2 minutes, then check the output:

kubectl get jobs -n jobs-demo -l job-name
kubectl logs -l job-name -n jobs-demo --prefix --tail=10

What is Happening

manifests/
  namespace.yaml      # jobs-demo namespace
  simple-job.yaml     # Single-pod compute task (pi digits)
  parallel-job.yaml   # 5 completions, 2 workers at a time
  failing-job.yaml    # Random failures with backoff and deadline
  cronjob.yaml        # Health report every 2 minutes

Key fields:

Field	Meaning
completions	Total number of successful pod runs needed
parallelism	Max pods running at the same time
backoffLimit	Max retries before marking the job as failed
activeDeadlineSeconds	Hard timeout for the entire job
schedule	Cron expression for CronJob timing
concurrencyPolicy	Forbid skips a run if the previous one is still active
successfulJobsHistoryLimit	How many completed Jobs to keep

Experiment

1. Trigger a CronJob run manually:

   kubectl create job manual-report --from=cronjob/health-reporter -n jobs-demo
   kubectl logs job/manual-report -n jobs-demo

2. Suspend a CronJob:

   kubectl patch cronjob health-reporter -n jobs-demo -p '{"spec":{"suspend":true}}'

3. Increase parallelism on the batch job:

   kubectl delete job batch-processor -n jobs-demo
   # Edit parallel-job.yaml: parallelism: 5
   kubectl apply -f manifests/parallel-job.yaml

Cleanup

kubectl delete namespace jobs-demo

Further Reading

See docs/deep-dive.md for a detailed explanation of Job controller internals, indexed jobs, TTL-after-finished, CronJob timezone handling, and real-world patterns like fan-out processing.

Next Step

Move on to DaemonSet to learn how to run a pod on every node.