Respect container exit code and job failure when using verify.executionMode.kubernetesCluster.jobManifestPath

[dwhoman]

This is a communication issue between Google Cloud Build verify and Skaffold verify. Skaffold allows you to specify a Kubernetes Job file.

apiVersion: skaffold/v4beta13
kind: Config
verify:
- name: "verify-fail"
  container:
    name: test-verifier
    image: ubuntu
  executionMode:
    kubernetesCluster:
      jobManifestPath: test-job.yaml

test-job.yaml is a Kubernetes job

apiVersion: batch/v1
kind: Job
metadata:
  name: test-verifier
  namespace: deployment-ns
spec:
  template:
    spec:
      containers:
      - name: test-verifier
        image: ubuntu
        command: ["/bin/sh"]
        args: ["-c", "echo 'Test fail'; exit 1"]
      restartPolicy: Never
  backoffLimit: 0
  podFailurePolicy:
    rules:
    - action: FailJob
      onExitCodes:
        containerName: test-verifier
        operator: NotIn
        values: [0]

The exit code is not respected; it does not cause the Google Cloud deploy verify to fail. I would suppose that Skaffold is not picking up the exit code of the test-verifier container and the subsequent failure of the test-verifier job. I would expect the behavior to be the same as

apiVersion: skaffold/v4beta13
kind: Config
verify:
- name: "verify-fail"
  container:
    name: test-verifier
    image: ubuntu
    command: ["/bin/sh"]
    args: ["-c", "echo 'Test fail'; exit 1" ]
  executionMode:
    kubernetesCluster: {}

but run in the default namespace rather than in the namespace specified in the Kubernetes Job definition. This version causes Google Cloud Deploy verify to fail as expected.