Introducing Indexed Jobs

Author: Aldo Culquicondor (Google)

Once you have containerized a non-parallel Job, it is quite easy to get it up and running on Kubernetes without modifications to the binary. In most cases, when running parallel distributed Jobs, you had to set a separate system to partition the work among the workers. For example, you could set up a task queue to assign one work item to each Pod or multiple items to each Pod until the queue is emptied.

The Kubernetes 1.21 release introduces a new field to control Job completion mode, a configuration option that allows you to control how Pod completions affect the overall progress of a Job, with two possible options (for now):

  • NonIndexed (default): the Job is considered complete when there has been a number of successfully completed Pods equal to the specified number in .spec.completions. In other words, each Pod completion is homologous to each other. Any Job you might have created before the introduction of completion modes is implicitly NonIndexed.
  • Indexed: the Job is considered complete when there is one successfully completed Pod associated with each index from 0 to .spec.completions-1. The index is exposed to each Pod in the annotation and the JOB_COMPLETION_INDEX environment variable.

You can start using Jobs with Indexed completion mode, or Indexed Jobs, for short, to easily start parallel Jobs. Then, each worker Pod can have a statically assigned partition of the data based on the index. This saves you from having to set up a queuing system or even having to modify your binary!

Creating an Indexed Job

To create an Indexed Job, you just have to add completionMode: Indexed to the Job spec and make use of the JOB_COMPLETION_INDEX environment variable.

apiVersion: batch/v1
kind: Job
  name: 'sample-job'
  completions: 3
  parallelism: 3
  completionMode: Indexed
      restartPolicy: Never
      - command:
        - 'bash'
        - '-c'
        - 'echo "My partition: ${JOB_COMPLETION_INDEX}"'
        image: ''
        name: 'sample-load'

Note that completion mode is an alpha feature in the 1.21 release. To be able to use it in your cluster, make sure to enable the IndexedJob feature gate on the API server and the controller manager.

When you run the example, you will see that each of the three created Pods gets a different completion index. For the user's convenience, the control plane sets the JOB_COMPLETION_INDEX environment variable, but you can choose to set your own or expose the index as a file.

See Indexed Job for parallel processing with static work assignment for a step-by-step guide, and a few more examples.

Future plans

SIG Apps envisions that there might be more completion modes that enable more use cases for the Job API. We welcome you to open issues in kubernetes/kubernetes with your suggestions.

In particular, we are considering an IndexedAndUnique mode where the indexes are not just available as annotation, but they are part of the Pod names, similar to StatefulSet. This should facilitate inter-Pod communication for tightly coupled Pods. You can join the discussion in the open issue.


Indexed Jobs allows you to statically partition work among the workers of your parallel Jobs. SIG Apps hopes that this feature facilitates the migration of more batch workloads to Kubernetes.