From 67cf648ab7701890f6b41804407b5a43f8eaf62f Mon Sep 17 00:00:00 2001
From: Maciej Szulik <maszulik@redhat.com>
Date: Mon, 8 Jan 2024 12:47:25 +0100
Subject: [PATCH 1/2] Add a new neverTerminate job behavior just for upgrade

---
 test/e2e/framework/job/fixtures.go | 6 ++++++
 test/e2e/upgrades/apps/job.go      | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/test/e2e/framework/job/fixtures.go b/test/e2e/framework/job/fixtures.go
index 27c46289043..6514bb3e243 100644
--- a/test/e2e/framework/job/fixtures.go
+++ b/test/e2e/framework/job/fixtures.go
@@ -92,6 +92,12 @@ func NewTestJobOnNode(behavior, name string, rPol v1.RestartPolicy, parallelism,
 		}
 	}
 	switch behavior {
+	case "neverTerminate":
+		// this job is being used in an upgrade job see test/e2e/upgrades/apps/job.go
+		// it should never be optimized, as it always has to restart during an upgrade
+		// and continue running
+		job.Spec.Template.Spec.Containers[0].Command = []string{"sleep", "1000000"}
+		job.Spec.Template.Spec.TerminationGracePeriodSeconds = ptr.To(int64(1))
 	case "notTerminate":
 		job.Spec.Template.Spec.Containers[0].Image = imageutils.GetPauseImageName()
 	case "fail":
diff --git a/test/e2e/upgrades/apps/job.go b/test/e2e/upgrades/apps/job.go
index eb59b4a8cdc..4c1b6e1b298 100644
--- a/test/e2e/upgrades/apps/job.go
+++ b/test/e2e/upgrades/apps/job.go
@@ -47,7 +47,7 @@ func (t *JobUpgradeTest) Setup(ctx context.Context, f *framework.Framework) {
 	t.namespace = f.Namespace.Name
 
 	ginkgo.By("Creating a job")
-	t.job = e2ejob.NewTestJob("notTerminate", "foo", v1.RestartPolicyOnFailure, 2, 2, nil, 6)
+	t.job = e2ejob.NewTestJob("neverTerminate", "foo", v1.RestartPolicyOnFailure, 2, 2, nil, 6)
 	job, err := e2ejob.CreateJob(ctx, f.ClientSet, t.namespace, t.job)
 	t.job = job
 	framework.ExpectNoError(err)

From f8abe71238f01d7e69c7eddcde1295d8751670ce Mon Sep 17 00:00:00 2001
From: Maciej Szulik <maszulik@redhat.com>
Date: Mon, 8 Jan 2024 13:52:03 +0100
Subject: [PATCH 2/2] Add an post-upgrade condition to ensure the job is
 running

---
 test/e2e/upgrades/apps/job.go | 23 +++++++++++++++++++++--
 1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/test/e2e/upgrades/apps/job.go b/test/e2e/upgrades/apps/job.go
index 4c1b6e1b298..b1311c82665 100644
--- a/test/e2e/upgrades/apps/job.go
+++ b/test/e2e/upgrades/apps/job.go
@@ -60,8 +60,11 @@ func (t *JobUpgradeTest) Setup(ctx context.Context, f *framework.Framework) {
 // Test verifies that the Jobs Pods are running after the an upgrade
 func (t *JobUpgradeTest) Test(ctx context.Context, f *framework.Framework, done <-chan struct{}, upgrade upgrades.UpgradeType) {
 	<-done
+	ginkgo.By("Ensuring job is running")
+	err := ensureJobRunning(ctx, f.ClientSet, t.namespace, t.job.Name)
+	framework.ExpectNoError(err)
 	ginkgo.By("Ensuring active pods == parallelism")
-	err := ensureAllJobPodsRunning(ctx, f.ClientSet, t.namespace, t.job.Name, 2)
+	err = ensureAllJobPodsRunning(ctx, f.ClientSet, t.namespace, t.job.Name, 2)
 	framework.ExpectNoError(err)
 }
 
@@ -70,7 +73,7 @@ func (t *JobUpgradeTest) Teardown(ctx context.Context, f *framework.Framework) {
 	// rely on the namespace deletion to clean up everything
 }
 
-// ensureAllJobPodsRunning uses c to check in the Job named jobName in ns
+// ensureAllJobPodsRunning uses c to check if the Job named jobName in ns
 // is running, returning an error if the expected parallelism is not
 // satisfied.
 func ensureAllJobPodsRunning(ctx context.Context, c clientset.Interface, ns, jobName string, parallelism int32) error {
@@ -93,3 +96,19 @@ func ensureAllJobPodsRunning(ctx context.Context, c clientset.Interface, ns, job
 	}
 	return nil
 }
+
+// ensureJobRunning uses c to check if the Job named jobName in ns is running,
+// (not completed, nor failed, nor suspended) returning an error if it can't
+// read the job or when it's not runnig
+func ensureJobRunning(ctx context.Context, c clientset.Interface, ns, jobName string) error {
+	job, err := e2ejob.GetJob(ctx, c, ns, jobName)
+	if err != nil {
+		return err
+	}
+	for _, c := range job.Status.Conditions {
+		if (c.Type == batchv1.JobComplete || c.Type == batchv1.JobFailed || c.Type == batchv1.JobSuspended) && c.Status == v1.ConditionTrue {
+			return fmt.Errorf("job is not running %#v", job)
+		}
+	}
+	return nil
+}