Users can also decide the number of tasks that should be finished in a single Job. This is helpful for solving some random sampling problems. Let's reuse the previous template and add spec.completions to see the differences:

$ cat repeat-job.yaml 
apiVersion: batch/v1
kind: Job
metadata:
  name: package-check-repeat
spec:
  activeDeadlineSeconds: 60
  completions: 3
  template:
    spec:
      containers:
      - name: package-check-repeat
        image: ubuntu
        command: ["dpkg-query", "-l"]
      restartPolicy: Never


$ kubectl create -f repeat-job.yaml 
job.batch/package-check-repeat created


$ kubectl get pods
NAME                         READY     STATUS        RESTARTS   AGE
package-check-7tfkt          0/1       Completed     0          52m
package-check-repeat-vl988   0/1       Completed     0          7s
package-check-repeat-flhz9   0/1       Completed     0          4s
package-check-repeat-xbf8b   0/1       Completed     0          2s

As you can see, three pods are created to complete this Job. This is useful if you need to run your program repeatedly at particular times. However, as you may have noticed from the Age column in the preceding result, these pods ran sequentially one by one. In the preceding result, the ages are 7 seconds, 4 seconds, then 2 seconds. This means that the second Job was started after the first Job was completed, and the third Job was started after the second Job was completed.

If it is the case that a Job runs for a longer period (such as a few days), but if there is no correlation needs between the 1^st, 2^nd, and 3^rd Job, then it does not make sense to run them sequentially. In this case, use a parallel Job.