Jobs

Jobs are immutable except activeDeadlineSeconds, parallelism and completions.

Cmds

k get jobs
kubectl create job example-job --image=busybox --dry-run=client -o yaml > job.yaml

Job Manifest

apiVersion: batch/v1
kind: Job
metadata:
  name: math-add-job
  labels:
    app: math
  annotations:
    description: "Performs simple parallel math tasks"
spec:
  # Clean up this Job object (and its pods) 60 seconds after it finishes
  ttlSecondsAfterFinished: 60

  # Number of successful completions needed
  completions: 3

  # Maximum number of pods to run in parallel
  parallelism: 3

  # Maximum number of pod restarts per failed pod before giving up
  backoffLimit: 4

  # Total execution time (in seconds) allowed for the Job (including retries)
  activeDeadlineSeconds: 120

  # Pod template
  template:
    spec:
      containers:
        - name: math-add-container
          image: busybox
          command: ["/bin/sh", "-c", "echo 2+2 = $((2+2))"]
      # For Jobs, restartPolicy must be OnFailure or Never
      restartPolicy: OnFailure

CronJob

Cmds

k get cronjob
k create cronjob throw-dice-cron-job --image kodekloud/throw-dice -o yaml --dry-run=client --schedule "30 21 * * *" > c.yaml

CronJob Manifest

apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: reporting-cron-job
  labels:
    app: reporting
  annotations:
    description: "This cron job runs a reporting task every minute."
spec:
  schedule: "*/1 * * * *" # every minute

  # Possible values: Allow (default), Forbid, or Replace
  concurrencyPolicy: Forbid # at the CronJob level

  # Retain N successful jobs for history
  successfulJobsHistoryLimit: 3

  # Retain N failed jobs for history
  failedJobsHistoryLimit: 2
  

  # If next scheduled time is missed by more than this number (seconds), skip it
  startingDeadlineSeconds: 60

  # If true, no new jobs are scheduled until set back to false
  suspend: false

  # ---- JOB TEMPLATE ----
  jobTemplate:
    metadata:
      labels:
        cronjob: "reporting"
    spec:
      # Maximum time (in seconds) a job can run before being terminated
      activeDeadlineSeconds: 300

      # Number of retries for a failing job
      backoffLimit: 6

      # Total completions needed
      completions: 2 

      # Number of pods to run in parallel
      # It works separately than concurrencyPolicy: Forbid
      parallelism: 2 # Job-level. how many Pods are created concurrently by a single Job

      # ---- POD SPEC (MINIMAL) ----
      template:
        spec:
          containers:
            - name: reporting-container
              image: your-reporting-image:latest
              # Simple example command
              command: ["/bin/sh", "-c", "echo Hello from the CronJob!"]
          # Typically OnFailure or Never for batch jobs
          restartPolicy: OnFailure

If parallelism is more than completions: Kubernetes will launch up to 5 pods at once, but only 2 successful completions are needed to mark the Job as complete.