completions and parallelism

Learn Job completions and parallelism fields.

completions and parallelism

A Job also manages tasks that must run many times. completions sets how many successful pods are needed in total, and parallelism how many pods run at most at the same time:

spec:
  completions: 10     # 10 successful runs
  parallelism: 3      # at most 3 at once

$ kubectl get job isle
NAME   STATUS    COMPLETIONS   DURATION
isle   Running   4/10          40s

With completionMode: Indexed, each pod gets an ordinal from 0 to 9 in the JOB_COMPLETION_INDEX environment variable and processes its own part of the work.

Indexed jobs

An indexed Job is used to split big data into parts and process them in parallel. Each pod picks a file or range according to its index:

spec:
  completions: 5
  parallelism: 5
  completionMode: Indexed
  template:
    spec:
      restartPolicy: Never
      containers:
        - name: isle
          image: busybox:1.36
          command: ["sh", "-c", "echo parca-$JOB_COMPLETION_INDEX isleniyor"]

Exercise Parallel processing

Write a Job's spec section (starting with spec:): completions: 5, parallelism: 5, completionMode: Indexed; restartPolicy: Never in the template; container named isle, image busybox:1.36.

Show example solution

spec:
  completions: 5
  parallelism: 5
  completionMode: Indexed
  template:
    spec:
      restartPolicy: Never
      containers:
        - name: isle
          image: busybox:1.36

Test yourself

With completions: 10, parallelism: 3, at least how many "waves" does the Job run?
- 1
- 4
- 3
- 10
Show answer

4 — 3 + 3 + 3 + 1 = 10.

Run at most 3 pods at the same time.

spec:
  completions: 10
  ____: 3

Show answer

spec:
  completions: 10
  parallelism: 3

Match.

completions

Required successes
parallelism

Concurrent pods
Indexed

Ordinal per pod
JOB_COMPLETION_INDEX

Index variable

Next lesson: What is a CronJob?

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies