Jobs

---

• Some tasks are short-lived and only need to run once and then finish.
• Examples include:
  • Generating thumbnail from image
  • Running data analytics
  • Generating a report and sending an email
• These are called batch workloads.

---

Jobs using Pod

• You create a Pod that performs the math operation.
• The pod starts, does the job, and exits but Kubernetes tries to restart the pod again and again. Why?
• Because of the restartPolicy, which is set to Always by default.

---

Stop from restarting It

• To prevent restart:
  • Set the pod’s restartPolicy to Never or OnFailure
  • Now the pod runs once and doesn’t restart after completion.
• But we still have a problem…

Multiple Pods for a Job

• Sometimes your task is large and needs to be ran across multiple pods.
• You want to run multiple pods and make sure all of them finish successfully,
• ReplicaSets won’t help here because they are meant to keep number of pods running all the time

---

That’s Where Jobs Come In

• A Job creates one or more pods and makes sure they finish the task

---

How to Create a Job

	apiVersion: batch/v1
	kind: Job
	metadata:
	  name: download-job
	spec:
	  template:
	    spec:
	      restartPolicy: Never
	      containers:
	        - name: downloader
	          image: curlimages/curl
	          command: ["curl", "-O", "https://example.com/data.csv"]

---

Running Multiple Pods

• To run more than one pod, set the completions field.
• For example:

  completions: 3

• This will run 3 pods, one after another.

---

What If Some Pods Fail?

• Kubernetes will keep creating new pods until it gets the desired number of successful completions.
• For example:
  • You ask for 3 completions.
  • If 2 pods succeed and 1 fails, Kubernetes tries again until 3 succeed total.

---

Running Pods in Parallel

• By default, job runs pods sequentially.

• To run them at the same time, add the parallelism field.

• For Example:

  parallelism: 3
  completions: 3

This means:

• Run 3 pods at once and If only 2 succeed, it creates 1 more pod and keeps doing this until 3 pods succeed

---

Commands

1. Create a job

kubectl create -f job.yaml

2. Check job status

kubectl get jobs

3. Get logs for job

kubectl logs <pod-name>

4. To delete the job

kubectl delete job <job-name>

---

⬅️ Blue Green and Canary Deployment | Cron Job ➡️

References

1. https://kubernetes.io/docs/concepts/workloads/controllers/job/