Kubernetes provides a activeDeadlineSeconds field for both JobSpec and PodSpec

What is the difference between the two? I've put together a little job with activeDeadlineSeconds set to 20, and in its Pod definition I have set the activeDeadlineSeconds field to 45. These are kind of arbitrary but meant to be spaced out. When I create/apply the Job then run kubectl get pods -a --watch, I can see that the 20 deadline isn't having any effect but the second one is (I see the DeadlineExceeded output).

Just to be extra certain, I added terminationGracePeriodSeconds: 10 in the PodSpec and see the same thing.

What is the purpose of the activeDeadlineSeconds in the Job? It doesn't seem to be sending any signal to my container.

Note: I'm simply running the sleep command on an ubuntu image. This command should exit when receiving the TERM signal sent by Kubernetes (so I expect a TERM signal at 20 seconds then the pod to die shortly thereafter)

Condensed YAML definition:

apiVersion: batch/v2alpha1  # K8s 1.7.x
kind: CronJob
spec:
  schedule: "*/1 * * * *"
  concurrencyPolicy: Allow
  jobTemplate:
    spec:  # JobSpec
      activeDeadlineSeconds: 20   # This needs to be shorter than the cron interval  ## TODO - NOT WORKING!
      parallelism: 1
      template:  # PodTemplateSpec
        spec:
          activeDeadlineSeconds: 45
          terminationGracePeriodSeconds: 10
          containers:
          - name: ubuntu-container
            image: ubuntu
            command: ['bash', '-c', 'sleep 500000']

References:

• https://v1-7.docs.kubernetes.io/docs/api-reference/v1.7/#jobspec-v1-batch
• https://v1-7.docs.kubernetes.io/docs/api-reference/v1.7/#podspec-v1-core
• https://unix.stackexchange.com/questions/429594/why-sigterm-is-able-to-kill-sleep