Job & CronJob - Batch Job

What is Job object in Kubernetes?

A Job object will be used to create one or more Pods and the Job ensures that a specified number of Pod instances will be created and terminates after completion of the Job. There could be finite jobs which will run within given certain timeout values. Job tracks for 'Successful' completion of the required task. Jobs can be run in two varients they can be parallel and also non-parallel.

Kubernetes Job types

There are 3 types of jobs non-parallel jobs [single pod jobs - unless it fails. creates replacement pod when pod goes down] parallel jobs with a fixed completion count parallel jobs with task queue 

##Example type 1: hundred-fibonaccis.yml

---
apiVersion: batch/v1
kind: Job
metadata:
    name: fibo-100
spec:
  template:
    spec: 
      containers:
      - name: fib-container 
        image: truek8s/hundred-fibonaccis:1.0
      restartPolicy: OnFailure
  backoffLimit: 3

Create the Job:

kubectl create -f hundred-fibonaccis.yml

Now let's describe the job:

kubectl describe job fibo-100

Describing a Job in Kubernetes

In the Job description you can observe the attributes such as parallelism, Pod Statuses.

 
On the other terminal which is running in parallel, Observe the pod status change from Running to Completed:

kubectl get po -w 

Pod still exists to get the logs

kubectl logs [podname] -- [container] 

We can see the Fibonacci number series printed out from the container logs.

Fibonacci series printed from kubctl logs command

##Example type 2

---
apiVersion: batch/v1
kind: Job
metadata:
  name: counter
spec:
  template:
    metadata:
      name: count-pod-template
    spec:
      containers:
      - name: count-container
        image: alpine
        command:
         - "/bin/sh"
         - "-c"
         - "for i in 100 50 10 5 1; do echo $i; done"
      restartPolicy: Never

To create the counter-job use the following command :

kubectl create -f counter-job.yaml

Check Job list:

kubectl get jobs

Check the Pod:

kubectl get po 

Check the log:

kubectl logs count-pod-xxx 

Counter Job execution output using kubectl logs

Now let's Describe the job counter :

kubectl describe jobs counter 

You can observer the Start Time and Pod Statuses from the above command ##cleanup delete job:

kubectl delete jobs counter-job 

no need to delete pod, When a Job deleted automatically removes the corresponding all its pods.

Controlling Job Completions

In some situations you need to run the same job multiple times, we need to define completions: number it under Job-> Specifications (spec section)

  ---
apiVersion: batch/v1
kind: Job
metadata:
  name: hello-job
spec:
  completions: 2
  template:
    spec:
      containers:
        - name: busybox-container
          image: busybox
          command: ["echo", "hello Kubernetes job!!!"]
       restartPolicy: Never
  

Observe the number of the Completions on the

watch kubectl get all
  kubectl get pods
  kubectl delete job hello-job
  kubectl get pods 

Once Job is deleted all its relavant resource will be cleaned up automaticall.

Parallelism

In some project there will be need to run multiple pods running in parallel

  ---
apiVersion: batch/v1
kind: Job
metadata:
  name: hello-job
spec:
  parallelism: 2
  template:
    spec:
      containers:
        - name: busybox-container
          image: busybox
          command: ["echo", "hello Kubernetes job!!!"]
       restartPolicy: Never
  

Observe the number of the Completions on the How backoffLimit works on Job? Let's do simple exeriment and understand this 'backoffLimit' attribute.

  ---
apiVersion: batch/v1
kind: Job
metadata:
  name: hello-job
spec:
  parallelism: 2
  backoffLiit: 4
  template:
    spec:
      containers:
        - name: busybox-container
          image: busybox
          command: ["ech0o", "hello Kubernetes job!!!"]
       restartPolicy: Never
  

Observe command mistyped purposefully, which will fail to create new Pod.

Working with CronJob

Job that works like a crontab in Linux systems. Any task that needs to be executed based on the scheduled time then we can use this Kubernetes Object.

##Example CronJob

---
apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: hello-cron
spec:
  schedule: "* * * * *"
  jobTemplate:
    spec:
      template:
        spec:
          containers:
          - name: busybox-container
            image: busybox
            command: ["echo", "Namste Kubernetes Cronjob!!!"]
            startPolicy: OnFailure

Creating

kubectl create -f cronjob.yaml

Open in new terminal and run the following command to watch continuously :

watch kubectl get all

Check the pods section in the above output