switch feature flag to beta for pod replacement policy and add e2e test

update pod replacement policy feature flag comment and refactor the e2e test for pod replacement policy

minor fixes for pod replacement policy and e2e test

fix wrong assertions for pod replacement policy e2e test

more fixes to pod replacement policy e2e test

refactor PodReplacementPolicy e2e test to use finalizers

fix unit tests when pod replacement policy feature flag is promoted to beta

fix podgc controller unit tests when pod replacement feature is enabled

fix lint issue in pod replacement policy e2e test

assert no error in defer function for removing finalizer in pod replacement policy e2e test

implement test using a sh trap for pod replacement policy

reduce sleep after SIGTERM in pod replacement policy e2e test to 5s

pkg/controller/job/job_controller_test.go

@@ -3452,6 +3452,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			wantStatus: batch.JobStatus{
 				Failed:                  1,
 				Succeeded:               2,
+				Terminating:             ptr.To[int32](0),
 				CompletedIndexes:        "0,1",
 				FailedIndexes:           ptr.To(""),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
@@ -3483,6 +3484,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			},
 			wantStatus: batch.JobStatus{
 				Active:                  2,
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 				FailedIndexes:           ptr.To(""),
 			},
@@ -3509,6 +3511,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			wantStatus: batch.JobStatus{
 				Active:                  2,
 				Failed:                  1,
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 				FailedIndexes:           ptr.To(""),
 			},
@@ -3535,6 +3538,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 				Active:                  1,
 				Failed:                  1,
 				FailedIndexes:           ptr.To("0"),
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 			},
 		},
@@ -3583,6 +3587,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 				Active:                  1,
 				Failed:                  1,
 				FailedIndexes:           ptr.To("0"),
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 			},
 		},
@@ -3632,6 +3637,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 				Active:                  0,
 				Failed:                  1,
 				FailedIndexes:           ptr.To(""),
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 				Conditions: []batch.JobCondition{
 					{
@@ -3695,6 +3701,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 				Active:                  2,
 				Failed:                  0,
 				FailedIndexes:           ptr.To(""),
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 			},
 		},
@@ -3721,6 +3728,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 				Failed:                  2,
 				Succeeded:               0,
 				FailedIndexes:           ptr.To(""),
+				Terminating:             ptr.To[int32](0),
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 				Conditions: []batch.JobCondition{
 					{
@@ -3754,6 +3762,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			wantStatus: batch.JobStatus{
 				Failed:                  1,
 				Succeeded:               1,
+				Terminating:             ptr.To[int32](0),
 				FailedIndexes:           ptr.To("0"),
 				CompletedIndexes:        "1",
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
@@ -3792,6 +3801,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			wantStatus: batch.JobStatus{
 				Failed:                  3,
 				Succeeded:               1,
+				Terminating:             ptr.To[int32](0),
 				FailedIndexes:           ptr.To("0,2"),
 				CompletedIndexes:        "1",
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
@@ -3830,6 +3840,7 @@ func TestSyncJobWithJobBackoffLimitPerIndex(t *testing.T) {
 			wantStatus: batch.JobStatus{
 				Active:                  2,
 				Succeeded:               1,
+				Terminating:             ptr.To[int32](0),
 				CompletedIndexes:        "1",
 				UncountedTerminatedPods: &batch.UncountedTerminatedPods{},
 			},

pkg/controller/podgc/gc_controller_test.go

@@ -295,6 +295,7 @@ func TestGCOrphaned(t *testing.T) {
 			},
 			itemsInQueue:    1,
 			deletedPodNames: sets.NewString("a"),
+			patchedPodNames: sets.NewString("a"),
 		},
 		{
 			name:                 "some nodes missing",
@@ -308,6 +309,7 @@ func TestGCOrphaned(t *testing.T) {
 			},
 			itemsInQueue:    1,
 			deletedPodNames: sets.NewString("a", "c", "d"),
+			patchedPodNames: sets.NewString("d"),
 		},
 		{
 			name:             "node added to client after quarantine",
@@ -457,6 +459,7 @@ func TestGCUnscheduledTerminating(t *testing.T) {
 				{name: "c", phase: v1.PodRunning, deletionTimeStamp: &metav1.Time{}, nodeName: ""},
 			},
 			deletedPodNames: sets.NewString("a", "b", "c"),
+			patchedPodNames: sets.NewString("c"),
 		},
 		{
 			name: "Scheduled pod in any phase must not be deleted",
@@ -607,6 +610,7 @@ func TestGCTerminating(t *testing.T) {
 				{name: "e6", phase: v1.PodUnknown, nodeName: "worker-5"},
 			},
 			deletedPodNames: sets.NewString("b1", "b4", "b5", "b6"),
+			patchedPodNames: sets.NewString("b1", "b4", "b5", "b6"),
 		},
 		{
 			name: "pods deleted from node tained out-of-service; PodDisruptionConditions enabled",

pkg/features/kube_features.go

@@ -389,6 +389,7 @@ const (
 	// owner: @kannon92
 	// kep : https://kep.k8s.io/3939
 	// alpha: v1.28
+	// beta: v1.29
 	//
 	// Allow users to specify recreating pods of a job only when
 	// pods have fully terminated.
@@ -990,7 +991,7 @@ var defaultKubernetesFeatureGates = map[featuregate.Feature]featuregate.FeatureS
 
 	JobPodFailurePolicy: {Default: true, PreRelease: featuregate.Beta},
 
-	JobPodReplacementPolicy: {Default: false, PreRelease: featuregate.Alpha},
+	JobPodReplacementPolicy: {Default: true, PreRelease: featuregate.Beta},
 
 	JobReadyPods: {Default: true, PreRelease: featuregate.GA, LockToDefault: true}, // remove in 1.31
 

test/e2e/apps/job.go

@@ -344,6 +344,61 @@ var _ = SIGDescribe("Job", func() {
 		}
 	})
 
+	ginkgo.It("should recreate pods only after they have failed if pod replacement policy is set to Failed", func(ctx context.Context) {
+		ginkgo.By("Creating a job")
+		job := e2ejob.NewTestJob("", "pod-recreate-failed", v1.RestartPolicyNever, 1, 1, nil, 1)
+		job.Spec.PodReplacementPolicy = ptr.To(batchv1.Failed)
+		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", `_term(){
+	sleep 5
+	exit 143
+}
+trap _term SIGTERM
+while true; do
+	sleep 1
+done`}
+		job, err := e2ejob.CreateJob(ctx, f.ClientSet, f.Namespace.Name, job)
+		framework.ExpectNoError(err, "failed to create job in namespace: %s", f.Namespace.Name)
+
+		err = e2ejob.WaitForJobPodsRunning(ctx, f.ClientSet, f.Namespace.Name, job.Name, 1)
+		framework.ExpectNoError(err, "failed to wait for job pod to become running in namespace: %s", f.Namespace.Name)
+
+		ginkgo.By("Deleting job pod")
+		pods, err := e2ejob.GetJobPods(ctx, f.ClientSet, f.Namespace.Name, job.Name)
+		framework.ExpectNoError(err, "failed to get pod list for job %s in namespace: %s", job.Name, f.Namespace.Name)
+
+		framework.ExpectNoError(e2epod.DeletePodsWithGracePeriod(ctx, f.ClientSet, pods.Items, 30), "failed to delete pods in namespace: %s", f.Namespace.Name)
+
+		ginkgo.By("Ensuring pod does not get recreated while it is in terminating state")
+		err = e2ejob.WaitForJobState(ctx, f.ClientSet, f.Namespace.Name, job.Name, f.Timeouts.PodDelete, func(job *batchv1.Job) string {
+			if job.Status.Active == 0 && job.Status.Failed == 0 && *job.Status.Terminating == 1 {
+				return ""
+			} else {
+				return fmt.Sprintf(
+					"expected job to have 0 active pod, 0 failed pod and 1 terminating pods, but got %d active pods, %d failed pods and %d terminating pods",
+					job.Status.Active,
+					job.Status.Failed,
+					*job.Status.Terminating,
+				)
+			}
+		})
+		framework.ExpectNoError(err, "failed to ensure pod is not recreated while it is in terminating state")
+
+		ginkgo.By("Ensuring pod gets recreated after it has failed")
+		err = e2ejob.WaitForJobState(ctx, f.ClientSet, f.Namespace.Name, job.Name, f.Timeouts.PodDelete, func(job *batchv1.Job) string {
+			if job.Status.Active == 1 && job.Status.Failed == 1 && *job.Status.Terminating == 0 {
+				return ""
+			} else {
+				return fmt.Sprintf(
+					"expected job to have 1 active pods, 1 failed pods and 0 terminating pod, but got %d active pods, %d failed pods and %d terminating pods",
+					job.Status.Active,
+					job.Status.Failed,
+					*job.Status.Terminating,
+				)
+			}
+		})
+		framework.ExpectNoError(err, "failed to wait for pod to get recreated")
+	})
+
 	/*
 		  Release: v1.24
 			Testname: Ensure Pods of an Indexed Job get a unique index.

test/integration/job/job_test.go

@@ -175,8 +175,9 @@ func TestJobPodFailurePolicyWithFailedPodDeletedDuringControllerRestart(t *testi
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, cs, jobObj, podsByStatus{
-		Active: count,
+		Active:      count,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	jobPods, err := getJobPods(ctx, t, cs, jobObj, func(s v1.PodStatus) bool {
@@ -452,8 +453,9 @@ func TestJobPodFailurePolicy(t *testing.T) {
 				t.Fatalf("Error %q while creating the job %q", err, jobObj.Name)
 			}
 			validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-				Active: 1,
+				Active:      1,
-				Ready:  ptr.To[int32](0),
+				Ready:       ptr.To[int32](0),
+				Terminating: ptr.To[int32](0),
 			})
 
 			op := func(p *v1.Pod) bool {
@@ -471,9 +473,10 @@ func TestJobPodFailurePolicy(t *testing.T) {
 			}
 
 			validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-				Active: test.wantActive,
+				Active:      test.wantActive,
-				Failed: test.wantFailed,
+				Failed:      test.wantFailed,
-				Ready:  ptr.To[int32](0),
+				Ready:       ptr.To[int32](0),
+				Terminating: ptr.To[int32](0),
 			})
 
 			if test.wantJobConditionType == batchv1.JobComplete {
@@ -517,8 +520,9 @@ func TestBackoffLimitPerIndex_DelayedPodDeletion(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0), "", ptr.To(""))
 
@@ -536,9 +540,10 @@ func TestBackoffLimitPerIndex_DelayedPodDeletion(t *testing.T) {
 	}
 
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Failed: 1,
+		Failed:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0), "", ptr.To(""))
 
@@ -559,10 +564,11 @@ func TestBackoffLimitPerIndex_DelayedPodDeletion(t *testing.T) {
 		t.Fatal("Failed trying to fail pod with index 0")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    0,
+		Active:      0,
-		Succeeded: 1,
+		Succeeded:   1,
-		Failed:    1,
+		Failed:      1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
 }
@@ -591,8 +597,9 @@ func TestBackoffLimitPerIndex_Reenabling(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 3,
+		Active:      3,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1, 2), "", ptr.To(""))
 
@@ -601,9 +608,10 @@ func TestBackoffLimitPerIndex_Reenabling(t *testing.T) {
 		t.Fatal("Failed trying to fail pod with index 0")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Failed: 1,
+		Failed:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(1, 2), "", ptr.To("0"))
 
@@ -615,9 +623,10 @@ func TestBackoffLimitPerIndex_Reenabling(t *testing.T) {
 		t.Fatal("Failed trying to fail pod with index 1")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 3,
+		Active:      3,
-		Failed: 2,
+		Failed:      2,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1, 2), "", nil)
 
@@ -632,9 +641,10 @@ func TestBackoffLimitPerIndex_Reenabling(t *testing.T) {
 	// Verify the indexes 0 and 1 are active as the failed pods don't have
 	// finalizers at this point, so they are ignored.
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Failed: 3,
+		Failed:      3,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1), "", ptr.To("2"))
 
@@ -676,8 +686,9 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Could not create job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1), "", ptr.To(""))
 
@@ -686,9 +697,10 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Failed: 1,
+		Failed:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1), "", ptr.To(""))
 
@@ -697,9 +709,10 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Failed: 2,
+		Failed:      2,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1), "", ptr.To(""))
 
@@ -708,9 +721,10 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Failed: 3,
+		Failed:      3,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1), "", ptr.To(""))
 
@@ -719,10 +733,11 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodSucceeded, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    1,
+		Active:      1,
-		Failed:    3,
+		Failed:      3,
-		Succeeded: 1,
+		Succeeded:   1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(1), "0", ptr.To(""))
 
@@ -731,10 +746,11 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    1,
+		Active:      1,
-		Failed:    4,
+		Failed:      4,
-		Succeeded: 1,
+		Succeeded:   1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(1), "0", ptr.To(""))
 
@@ -743,10 +759,11 @@ func TestBackoffLimitPerIndex_JobPodsCreatedWithExponentialBackoff(t *testing.T)
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodSucceeded, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    0,
+		Active:      0,
-		Failed:    4,
+		Failed:      4,
-		Succeeded: 2,
+		Succeeded:   2,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New[int](), "0,1", ptr.To(""))
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
@@ -1106,8 +1123,9 @@ func TestBackoffLimitPerIndex(t *testing.T) {
 				t.Fatalf("Error %q while creating the job %q", err, jobObj.Name)
 			}
 			validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-				Active: int(*test.job.Spec.Parallelism),
+				Active:      int(*test.job.Spec.Parallelism),
-				Ready:  ptr.To[int32](0),
+				Ready:       ptr.To[int32](0),
+				Terminating: ptr.To[int32](0),
 			})
 			for _, podTermination := range test.podTerminations {
 				pod, err := getActivePodForIndex(ctx, clientSet, jobObj, podTermination.index)
@@ -1119,10 +1137,11 @@ func TestBackoffLimitPerIndex(t *testing.T) {
 					t.Fatalf("Error updating the pod %q: %q", klog.KObj(pod), err)
 				}
 				validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-					Active:    podTermination.wantActive,
+					Active:      podTermination.wantActive,
-					Succeeded: podTermination.wantSucceeded,
+					Succeeded:   podTermination.wantSucceeded,
-					Failed:    podTermination.wantFailed,
+					Failed:      podTermination.wantFailed,
-					Ready:     ptr.To[int32](0),
+					Ready:       ptr.To[int32](0),
+					Terminating: ptr.To[int32](0),
 				})
 				validateIndexedJobPods(ctx, t, clientSet, jobObj, podTermination.wantActiveIndexes, podTermination.wantCompletedIndexes, podTermination.wantFailedIndexes)
 				if podTermination.wantReplacementPodFailureCount != nil {
@@ -1187,8 +1206,9 @@ func TestNonParallelJob(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Restarting controller.
@@ -1200,9 +1220,10 @@ func TestNonParallelJob(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Failed: 1,
+		Failed:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateCounterMetric(ctx, t, metrics.JobPodsFinished, metricLabelsWithValue{
 		Labels: []string{"NonIndexed", "failed"},
@@ -1219,9 +1240,10 @@ func TestNonParallelJob(t *testing.T) {
 	}
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Failed:    1,
+		Failed:      1,
-		Succeeded: 1,
+		Succeeded:   1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateFinishedPodsNoFinalizer(ctx, t, clientSet, jobObj)
 	validateCounterMetric(ctx, t, metrics.JobFinishedNum, metricLabelsWithValue{
@@ -1251,8 +1273,9 @@ func TestParallelJob(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	want := podsByStatus{
-		Active: 5,
+		Active:      5,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 
@@ -1268,9 +1291,10 @@ func TestParallelJob(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pods: %v", v1.PodFailed, err)
 	}
 	want = podsByStatus{
-		Active: 5,
+		Active:      5,
-		Failed: 2,
+		Failed:      2,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	// Once one Pod succeeds, no more Pods are created, even if some fail.
@@ -1278,20 +1302,22 @@ func TestParallelJob(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodSucceeded, err)
 	}
 	want = podsByStatus{
-		Failed:    2,
+		Failed:      2,
-		Succeeded: 1,
+		Succeeded:   1,
-		Active:    4,
+		Active:      4,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	if err, _ := setJobPodsPhase(ctx, clientSet, jobObj, v1.PodFailed, 2); err != nil {
 		t.Fatalf("Failed setting phase %s on Job Pods: %v", v1.PodFailed, err)
 	}
 	want = podsByStatus{
-		Failed:    4,
+		Failed:      4,
-		Succeeded: 1,
+		Succeeded:   1,
-		Active:    2,
+		Active:      2,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	// No more Pods are created after remaining Pods succeed.
@@ -1300,9 +1326,10 @@ func TestParallelJob(t *testing.T) {
 	}
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
 	want = podsByStatus{
-		Failed:    4,
+		Failed:      4,
-		Succeeded: 3,
+		Succeeded:   3,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	validateFinishedPodsNoFinalizer(ctx, t, clientSet, jobObj)
@@ -1337,8 +1364,9 @@ func TestParallelJobChangingParallelism(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 5,
+		Active:      5,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Reduce parallelism by a number greater than backoffLimit.
@@ -1348,8 +1376,9 @@ func TestParallelJobChangingParallelism(t *testing.T) {
 		t.Fatalf("Updating Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 2,
+		Active:      2,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Increase parallelism again.
@@ -1359,8 +1388,9 @@ func TestParallelJobChangingParallelism(t *testing.T) {
 		t.Fatalf("Updating Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 4,
+		Active:      4,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Succeed Job
@@ -1369,8 +1399,9 @@ func TestParallelJobChangingParallelism(t *testing.T) {
 	}
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Succeeded: 4,
+		Succeeded:   4,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateFinishedPodsNoFinalizer(ctx, t, clientSet, jobObj)
 }
@@ -1397,8 +1428,9 @@ func TestParallelJobWithCompletions(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	want := podsByStatus{
-		Active: 54,
+		Active:      54,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	// Tracks ready pods, if enabled.
@@ -1413,9 +1445,10 @@ func TestParallelJobWithCompletions(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pods: %v", v1.PodFailed, err)
 	}
 	want = podsByStatus{
-		Active: 54,
+		Active:      54,
-		Failed: 2,
+		Failed:      2,
-		Ready:  ptr.To[int32](50),
+		Ready:       ptr.To[int32](50),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	// Pods are created until the number of succeeded Pods equals completions.
@@ -1423,10 +1456,11 @@ func TestParallelJobWithCompletions(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodSucceeded, err)
 	}
 	want = podsByStatus{
-		Failed:    2,
+		Failed:      2,
-		Succeeded: 53,
+		Succeeded:   53,
-		Active:    3,
+		Active:      3,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	// No more Pods are created after the Job completes.
@@ -1435,9 +1469,10 @@ func TestParallelJobWithCompletions(t *testing.T) {
 	}
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
 	want = podsByStatus{
-		Failed:    2,
+		Failed:      2,
-		Succeeded: 56,
+		Succeeded:   56,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, want)
 	validateFinishedPodsNoFinalizer(ctx, t, clientSet, jobObj)
@@ -1475,8 +1510,9 @@ func TestIndexedJob(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 3,
+		Active:      3,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 1, 2), "", nil)
 	validateCounterMetric(ctx, t, metrics.JobFinishedIndexesTotal, metricLabelsWithValue{
@@ -1489,9 +1525,10 @@ func TestIndexedJob(t *testing.T) {
 		t.Fatal("Failed trying to succeed pod with index 1")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    3,
+		Active:      3,
-		Succeeded: 1,
+		Succeeded:   1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 2, 3), "1", nil)
 	validateCounterMetric(ctx, t, metrics.JobFinishedIndexesTotal, metricLabelsWithValue{
@@ -1504,10 +1541,11 @@ func TestIndexedJob(t *testing.T) {
 		t.Fatal("Failed trying to succeed pod with index 2")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    3,
+		Active:      3,
-		Failed:    1,
+		Failed:      1,
-		Succeeded: 1,
+		Succeeded:   1,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, sets.New(0, 2, 3), "1", nil)
 	validateCounterMetric(ctx, t, metrics.JobFinishedIndexesTotal, metricLabelsWithValue{
@@ -1520,10 +1558,11 @@ func TestIndexedJob(t *testing.T) {
 		t.Fatal("Failed trying to succeed remaining pods")
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active:    0,
+		Active:      0,
-		Failed:    1,
+		Failed:      1,
-		Succeeded: 4,
+		Succeeded:   4,
-		Ready:     ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	validateIndexedJobPods(ctx, t, clientSet, jobObj, nil, "0-3", nil)
 	validateJobSucceeded(ctx, t, clientSet, jobObj)
@@ -2028,10 +2067,11 @@ func TestElasticIndexedJob(t *testing.T) {
 				}
 
 				validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-					Active:    update.wantActivePods,
+					Active:      update.wantActivePods,
-					Succeeded: len(update.succeedIndexes),
+					Succeeded:   len(update.succeedIndexes),
-					Failed:    update.wantFailed,
+					Failed:      update.wantFailed,
-					Ready:     ptr.To[int32](0),
+					Ready:       ptr.To[int32](0),
+					Terminating: ptr.To[int32](0),
 				})
 				validateIndexedJobPods(ctx, t, clientSet, jobObj, update.wantRemainingIndexes, update.wantSucceededIndexes, nil)
 			}
@@ -2147,8 +2187,9 @@ func TestOrphanPodsFinalizersClearedWithGC(t *testing.T) {
 				t.Fatalf("Failed to create Job: %v", err)
 			}
 			validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-				Active: 2,
+				Active:      2,
-				Ready:  ptr.To[int32](0),
+				Ready:       ptr.To[int32](0),
+				Terminating: ptr.To[int32](0),
 			})
 
 			// Delete Job. The GC should delete the pods in cascade.
@@ -2222,8 +2263,9 @@ func TestJobPodsCreatedWithExponentialBackoff(t *testing.T) {
 		t.Fatalf("Could not create job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Fail the first pod
@@ -2231,9 +2273,10 @@ func TestJobPodsCreatedWithExponentialBackoff(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
-		Failed: 1,
+		Failed:      1,
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Fail the second pod
@@ -2241,9 +2284,10 @@ func TestJobPodsCreatedWithExponentialBackoff(t *testing.T) {
 		t.Fatalf("Failed setting phase %s on Job Pod: %v", v1.PodFailed, err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
-		Failed: 2,
+		Failed:      2,
+		Terminating: ptr.To[int32](0),
 	})
 
 	jobPods, err := getJobPods(ctx, t, clientSet, jobObj, func(ps v1.PodStatus) bool { return true })
@@ -2320,8 +2364,9 @@ func TestJobFailedWithInterrupts(t *testing.T) {
 		t.Fatalf("Could not create job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 10,
+		Active:      10,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 	t.Log("Finishing pods")
 	if err, _ := setJobPodsPhase(ctx, clientSet, jobObj, v1.PodFailed, 1); err != nil {
@@ -2385,8 +2430,9 @@ func TestOrphanPodsFinalizersClearedOnRestart(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, jobObj, podsByStatus{
-		Active: 1,
+		Active:      1,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 
 	// Step 2: Delete the Job while the controller is stopped.
@@ -2453,8 +2499,9 @@ func TestSuspendJob(t *testing.T) {
 
 			validate := func(s string, active int, status v1.ConditionStatus, reason string) {
 				validateJobPodsStatus(ctx, t, clientSet, job, podsByStatus{
-					Active: active,
+					Active:      active,
-					Ready:  ptr.To[int32](0),
+					Ready:       ptr.To[int32](0),
+					Terminating: ptr.To[int32](0),
 				})
 				job, err = clientSet.BatchV1().Jobs(ns.Name).Get(ctx, job.Name, metav1.GetOptions{})
 				if err != nil {
@@ -2496,8 +2543,9 @@ func TestSuspendJobControllerRestart(t *testing.T) {
 		t.Fatalf("Failed to create Job: %v", err)
 	}
 	validateJobPodsStatus(ctx, t, clientSet, job, podsByStatus{
-		Active: 0,
+		Active:      0,
-		Ready:  ptr.To[int32](0),
+		Ready:       ptr.To[int32](0),
+		Terminating: ptr.To[int32](0),
 	})
 }