Keep failed CronJob Pods around in Kubernetes

There are a few fields that play into this, but the gist of it is to set the Pod's restartPolicy to Never. This doesn't mean that the Job only tries once. Rather, instead of restarting the container inside the Pod on failure, a new Pod will get scheduled until the Job's backoffLimit is reached (default being 6 at the moment). Finally the CronJob's failedJobsHistoryLimit decides whether the failed Job together with its Pods will be kept around. It defaults to 1.

Here's an example of a debuggable CronJob, adapted from the official example (CC BY 4.0, The Kubernetes Authors):

apiVersion: batch/v1
kind: CronJob
metadata:
  name: hello
spec:
  schedule: "* * * * *"
  jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: hello
              image: busybox:1.28
              imagePullPolicy: IfNotPresent
              command:
                - /bin/sh
                - -c
                - date; echo Hello from the Kubernetes cluster
            # This is the key bit.
            # Defaults are okay for the other discussed fields.
            restartPolicy: Never

posted 6 days ago

CC BY-SA 4.0

Iizuki‭

1297 reputation 26 28 168 30