diff --git a/test/e2e/apps/BUILD b/test/e2e/apps/BUILD
index 3857cea166a..90a4cf991d8 100644
--- a/test/e2e/apps/BUILD
+++ b/test/e2e/apps/BUILD
@@ -61,6 +61,7 @@ go_library(
         "//staging/src/k8s.io/client-go/tools/watch:go_default_library",
         "//test/e2e/common:go_default_library",
         "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/e2e/framework/replicaset:go_default_library",
         "//test/utils:go_default_library",
         "//test/utils/image:go_default_library",
diff --git a/test/e2e/apps/cronjob.go b/test/e2e/apps/cronjob.go
index 20bb632d70e..12af593d611 100644
--- a/test/e2e/apps/cronjob.go
+++ b/test/e2e/apps/cronjob.go
@@ -34,6 +34,7 @@ import (
 	batchinternal "k8s.io/kubernetes/pkg/apis/batch"
 	"k8s.io/kubernetes/pkg/controller/job"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 	imageutils "k8s.io/kubernetes/test/utils/image"
 )
 
@@ -210,7 +211,7 @@ var _ = SIGDescribe("CronJob", func() {
 		framework.ExpectNoError(framework.DeleteResourceAndWaitForGC(f.ClientSet, batchinternal.Kind("Job"), f.Namespace.Name, job.Name))
 
 		By("Ensuring job was deleted")
-		_, err = framework.GetJob(f.ClientSet, f.Namespace.Name, job.Name)
+		_, err = jobutil.GetJob(f.ClientSet, f.Namespace.Name, job.Name)
 		Expect(err).To(HaveOccurred())
 		Expect(errors.IsNotFound(err)).To(BeTrue())
 
diff --git a/test/e2e/apps/job.go b/test/e2e/apps/job.go
index 620b1677850..dbb8f2ba83c 100644
--- a/test/e2e/apps/job.go
+++ b/test/e2e/apps/job.go
@@ -25,6 +25,7 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	batchinternal "k8s.io/kubernetes/pkg/apis/batch"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 
 	. "github.com/onsi/ginkgo"
 	. "github.com/onsi/gomega"
@@ -39,12 +40,12 @@ var _ = SIGDescribe("Job", func() {
 	// Simplest case: all pods succeed promptly
 	It("should run a job to completion when tasks succeed", func() {
 		By("Creating a job")
-		job := framework.NewTestJob("succeed", "all-succeed", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("succeed", "all-succeed", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 
 		By("Ensuring job reaches completions")
-		err = framework.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, completions)
+		err = jobutil.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, completions)
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job completion in namespace: %s", f.Namespace.Name)
 	})
 
@@ -58,12 +59,12 @@ var _ = SIGDescribe("Job", func() {
 		// up to 5 minutes between restarts, making test timeouts
 		// due to successive failures too likely with a reasonable
 		// test timeout.
-		job := framework.NewTestJob("failOnce", "fail-once-local", v1.RestartPolicyOnFailure, parallelism, completions, nil, backoffLimit)
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("failOnce", "fail-once-local", v1.RestartPolicyOnFailure, parallelism, completions, nil, backoffLimit)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 
 		By("Ensuring job reaches completions")
-		err = framework.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, completions)
+		err = jobutil.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, completions)
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job completion in namespace: %s", f.Namespace.Name)
 	})
 
@@ -79,61 +80,61 @@ var _ = SIGDescribe("Job", func() {
 		// With the introduction of backoff limit and high failure rate this
 		// is hitting its timeout, the 3 is a reasonable that should make this
 		// test less flaky, for now.
-		job := framework.NewTestJob("randomlySucceedOrFail", "rand-non-local", v1.RestartPolicyNever, parallelism, 3, nil, 999)
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("randomlySucceedOrFail", "rand-non-local", v1.RestartPolicyNever, parallelism, 3, nil, 999)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 
 		By("Ensuring job reaches completions")
-		err = framework.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, *job.Spec.Completions)
+		err = jobutil.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, *job.Spec.Completions)
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job completion in namespace: %s", f.Namespace.Name)
 	})
 
 	It("should exceed active deadline", func() {
 		By("Creating a job")
 		var activeDeadlineSeconds int64 = 1
-		job := framework.NewTestJob("notTerminate", "exceed-active-deadline", v1.RestartPolicyNever, parallelism, completions, &activeDeadlineSeconds, backoffLimit)
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("notTerminate", "exceed-active-deadline", v1.RestartPolicyNever, parallelism, completions, &activeDeadlineSeconds, backoffLimit)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 		By("Ensuring job past active deadline")
-		err = framework.WaitForJobFailure(f.ClientSet, f.Namespace.Name, job.Name, time.Duration(activeDeadlineSeconds+10)*time.Second, "DeadlineExceeded")
+		err = jobutil.WaitForJobFailure(f.ClientSet, f.Namespace.Name, job.Name, time.Duration(activeDeadlineSeconds+10)*time.Second, "DeadlineExceeded")
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job past active deadline in namespace: %s", f.Namespace.Name)
 	})
 
 	It("should delete a job", func() {
 		By("Creating a job")
-		job := framework.NewTestJob("notTerminate", "foo", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("notTerminate", "foo", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 
 		By("Ensuring active pods == parallelism")
-		err = framework.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, job.Name, parallelism)
+		err = jobutil.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, job.Name, parallelism)
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure active pods == parallelism in namespace: %s", f.Namespace.Name)
 
 		By("delete a job")
 		framework.ExpectNoError(framework.DeleteResourceAndWaitForGC(f.ClientSet, batchinternal.Kind("Job"), f.Namespace.Name, job.Name))
 
 		By("Ensuring job was deleted")
-		_, err = framework.GetJob(f.ClientSet, f.Namespace.Name, job.Name)
+		_, err = jobutil.GetJob(f.ClientSet, f.Namespace.Name, job.Name)
 		Expect(err).To(HaveOccurred(), "failed to ensure job %s was deleted in namespace: %s", job.Name, f.Namespace.Name)
 		Expect(errors.IsNotFound(err)).To(BeTrue())
 	})
 
 	It("should adopt matching orphans and release non-matching pods", func() {
 		By("Creating a job")
-		job := framework.NewTestJob("notTerminate", "adopt-release", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
+		job := jobutil.NewTestJob("notTerminate", "adopt-release", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
 		// Replace job with the one returned from Create() so it has the UID.
 		// Save Kind since it won't be populated in the returned job.
 		kind := job.Kind
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 		job.Kind = kind
 
 		By("Ensuring active pods == parallelism")
-		err = framework.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, job.Name, parallelism)
+		err = jobutil.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, job.Name, parallelism)
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure active pods == parallelism in namespace: %s", f.Namespace.Name)
 
 		By("Orphaning one of the Job's Pods")
-		pods, err := framework.GetJobPods(f.ClientSet, f.Namespace.Name, job.Name)
+		pods, err := jobutil.GetJobPods(f.ClientSet, f.Namespace.Name, job.Name)
 		Expect(err).NotTo(HaveOccurred(), "failed to get PodList for job %s in namespace: %s", job.Name, f.Namespace.Name)
 		Expect(pods.Items).To(HaveLen(int(parallelism)))
 		pod := pods.Items[0]
@@ -142,7 +143,7 @@ var _ = SIGDescribe("Job", func() {
 		})
 
 		By("Checking that the Job readopts the Pod")
-		Expect(framework.WaitForPodCondition(f.ClientSet, pod.Namespace, pod.Name, "adopted", framework.JobTimeout,
+		Expect(framework.WaitForPodCondition(f.ClientSet, pod.Namespace, pod.Name, "adopted", jobutil.JobTimeout,
 			func(pod *v1.Pod) (bool, error) {
 				controllerRef := metav1.GetControllerOf(pod)
 				if controllerRef == nil {
@@ -161,7 +162,7 @@ var _ = SIGDescribe("Job", func() {
 		})
 
 		By("Checking that the Job releases the Pod")
-		Expect(framework.WaitForPodCondition(f.ClientSet, pod.Namespace, pod.Name, "released", framework.JobTimeout,
+		Expect(framework.WaitForPodCondition(f.ClientSet, pod.Namespace, pod.Name, "released", jobutil.JobTimeout,
 			func(pod *v1.Pod) (bool, error) {
 				controllerRef := metav1.GetControllerOf(pod)
 				if controllerRef != nil {
@@ -175,16 +176,16 @@ var _ = SIGDescribe("Job", func() {
 	It("should exceed backoffLimit", func() {
 		By("Creating a job")
 		backoff := 1
-		job := framework.NewTestJob("fail", "backofflimit", v1.RestartPolicyNever, 1, 1, nil, int32(backoff))
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job := jobutil.NewTestJob("fail", "backofflimit", v1.RestartPolicyNever, 1, 1, nil, int32(backoff))
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job in namespace: %s", f.Namespace.Name)
 		By("Ensuring job exceed backofflimit")
 
-		err = framework.WaitForJobFailure(f.ClientSet, f.Namespace.Name, job.Name, framework.JobTimeout, "BackoffLimitExceeded")
+		err = jobutil.WaitForJobFailure(f.ClientSet, f.Namespace.Name, job.Name, jobutil.JobTimeout, "BackoffLimitExceeded")
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job exceed backofflimit in namespace: %s", f.Namespace.Name)
 
 		By(fmt.Sprintf("Checking that %d pod created and status is failed", backoff+1))
-		pods, err := framework.GetJobPods(f.ClientSet, f.Namespace.Name, job.Name)
+		pods, err := jobutil.GetJobPods(f.ClientSet, f.Namespace.Name, job.Name)
 		Expect(err).NotTo(HaveOccurred(), "failed to get PodList for job %s in namespace: %s", job.Name, f.Namespace.Name)
 		// Expect(pods.Items).To(HaveLen(backoff + 1))
 		// due to NumRequeus not being stable enough, especially with failed status
diff --git a/test/e2e/apps/network_partition.go b/test/e2e/apps/network_partition.go
index 499d13cf098..97089f26619 100644
--- a/test/e2e/apps/network_partition.go
+++ b/test/e2e/apps/network_partition.go
@@ -36,6 +36,7 @@ import (
 	nodepkg "k8s.io/kubernetes/pkg/controller/nodelifecycle"
 	"k8s.io/kubernetes/test/e2e/common"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 	testutils "k8s.io/kubernetes/test/utils"
 
 	. "github.com/onsi/ginkgo"
@@ -426,11 +427,11 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 			completions := int32(4)
 			backoffLimit := int32(6) // default value
 
-			job := framework.NewTestJob("notTerminate", "network-partition", v1.RestartPolicyNever,
+			job := jobutil.NewTestJob("notTerminate", "network-partition", v1.RestartPolicyNever,
 				parallelism, completions, nil, backoffLimit)
-			job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+			job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 			Expect(err).NotTo(HaveOccurred())
-			label := labels.SelectorFromSet(labels.Set(map[string]string{framework.JobSelectorKey: job.Name}))
+			label := labels.SelectorFromSet(labels.Set(map[string]string{jobutil.JobSelectorKey: job.Name}))
 
 			By(fmt.Sprintf("verifying that there are now %v running pods", parallelism))
 			_, err = framework.PodsCreatedByLabel(c, ns, job.Name, parallelism, label)
diff --git a/test/e2e/auth/BUILD b/test/e2e/auth/BUILD
index 31689891c49..30ae4a39ebf 100644
--- a/test/e2e/auth/BUILD
+++ b/test/e2e/auth/BUILD
@@ -53,6 +53,7 @@ go_library(
         "//staging/src/k8s.io/client-go/util/cert:go_default_library",
         "//test/e2e/common:go_default_library",
         "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/utils:go_default_library",
         "//test/utils/image:go_default_library",
         "//vendor/github.com/evanphx/json-patch:go_default_library",
diff --git a/test/e2e/auth/metadata_concealment.go b/test/e2e/auth/metadata_concealment.go
index 58337dc8d21..3002b964e4c 100644
--- a/test/e2e/auth/metadata_concealment.go
+++ b/test/e2e/auth/metadata_concealment.go
@@ -21,6 +21,7 @@ import (
 	"k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 
 	. "github.com/onsi/ginkgo"
 	. "github.com/onsi/gomega"
@@ -54,11 +55,11 @@ var _ = SIGDescribe("Metadata Concealment", func() {
 				},
 			},
 		}
-		job, err := framework.CreateJob(f.ClientSet, f.Namespace.Name, job)
+		job, err := jobutil.CreateJob(f.ClientSet, f.Namespace.Name, job)
 		Expect(err).NotTo(HaveOccurred(), "failed to create job (%s:%s)", f.Namespace.Name, job.Name)
 
 		By("Ensuring job reaches completions")
-		err = framework.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, int32(1))
+		err = jobutil.WaitForJobComplete(f.ClientSet, f.Namespace.Name, job.Name, int32(1))
 		Expect(err).NotTo(HaveOccurred(), "failed to ensure job completion (%s:%s)", f.Namespace.Name, job.Name)
 	})
 })
diff --git a/test/e2e/framework/BUILD b/test/e2e/framework/BUILD
index db0f3be371f..054e1bdf804 100644
--- a/test/e2e/framework/BUILD
+++ b/test/e2e/framework/BUILD
@@ -14,7 +14,6 @@ go_library(
         "framework.go",
         "get-kubemark-resource-usage.go",
         "google_compute.go",
-        "jobs_util.go",
         "kubelet_stats.go",
         "log_size_monitoring.go",
         "metrics_util.go",
@@ -47,7 +46,6 @@ go_library(
         "//pkg/client/conditions:go_default_library",
         "//pkg/controller:go_default_library",
         "//pkg/controller/deployment/util:go_default_library",
-        "//pkg/controller/job:go_default_library",
         "//pkg/controller/nodelifecycle:go_default_library",
         "//pkg/controller/service:go_default_library",
         "//pkg/features:go_default_library",
@@ -155,6 +153,7 @@ filegroup(
         "//test/e2e/framework/ginkgowrapper:all-srcs",
         "//test/e2e/framework/gpu:all-srcs",
         "//test/e2e/framework/ingress:all-srcs",
+        "//test/e2e/framework/job:all-srcs",
         "//test/e2e/framework/metrics:all-srcs",
         "//test/e2e/framework/podlogs:all-srcs",
         "//test/e2e/framework/providers/aws:all-srcs",
diff --git a/test/e2e/framework/job/BUILD b/test/e2e/framework/job/BUILD
new file mode 100644
index 00000000000..e799bd082a8
--- /dev/null
+++ b/test/e2e/framework/job/BUILD
@@ -0,0 +1,38 @@
+load("@io_bazel_rules_go//go:def.bzl", "go_library")
+
+go_library(
+    name = "go_default_library",
+    srcs = [
+        "const.go",
+        "fixtures.go",
+        "rest.go",
+        "wait.go",
+    ],
+    importpath = "k8s.io/kubernetes/test/e2e/framework/job",
+    visibility = ["//visibility:public"],
+    deps = [
+        "//pkg/controller/job:go_default_library",
+        "//staging/src/k8s.io/api/batch/v1:go_default_library",
+        "//staging/src/k8s.io/api/core/v1:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/api/errors:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/labels:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
+        "//staging/src/k8s.io/client-go/kubernetes:go_default_library",
+        "//test/e2e/framework:go_default_library",
+    ],
+)
+
+filegroup(
+    name = "package-srcs",
+    srcs = glob(["**"]),
+    tags = ["automanaged"],
+    visibility = ["//visibility:private"],
+)
+
+filegroup(
+    name = "all-srcs",
+    srcs = [":package-srcs"],
+    tags = ["automanaged"],
+    visibility = ["//visibility:public"],
+)
diff --git a/test/e2e/framework/job/const.go b/test/e2e/framework/job/const.go
new file mode 100644
index 00000000000..e3bd80cc806
--- /dev/null
+++ b/test/e2e/framework/job/const.go
@@ -0,0 +1,27 @@
+/*
+Copyright 2019 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package job
+
+import "time"
+
+const (
+	// JobTimeout is how long to wait for a job to finish.
+	JobTimeout = 15 * time.Minute
+
+	// JobSelectorKey is a job selector name
+	JobSelectorKey = "job"
+)
diff --git a/test/e2e/framework/job/fixtures.go b/test/e2e/framework/job/fixtures.go
new file mode 100644
index 00000000000..f3caae4f706
--- /dev/null
+++ b/test/e2e/framework/job/fixtures.go
@@ -0,0 +1,109 @@
+/*
+Copyright 2019 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package job
+
+import (
+	batchv1 "k8s.io/api/batch/v1"
+	"k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// NewTestJob returns a Job which does one of several testing behaviors. notTerminate starts a Job that will run
+// effectively forever. fail starts a Job that will fail immediately. succeed starts a Job that will succeed
+// immediately. randomlySucceedOrFail starts a Job that will succeed or fail randomly. failOnce fails the Job the
+// first time it is run and succeeds subsequently. name is the Name of the Job. RestartPolicy indicates the restart
+// policy of the containers in which the Pod is running. Parallelism is the Job's parallelism, and completions is the
+// Job's required number of completions.
+func NewTestJob(behavior, name string, rPol v1.RestartPolicy, parallelism, completions int32, activeDeadlineSeconds *int64, backoffLimit int32) *batchv1.Job {
+	manualSelector := false
+	job := &batchv1.Job{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		TypeMeta: metav1.TypeMeta{
+			Kind: "Job",
+		},
+		Spec: batchv1.JobSpec{
+			ActiveDeadlineSeconds: activeDeadlineSeconds,
+			Parallelism:           &parallelism,
+			Completions:           &completions,
+			BackoffLimit:          &backoffLimit,
+			ManualSelector:        &manualSelector,
+			Template: v1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{JobSelectorKey: name},
+				},
+				Spec: v1.PodSpec{
+					RestartPolicy: rPol,
+					Volumes: []v1.Volume{
+						{
+							Name: "data",
+							VolumeSource: v1.VolumeSource{
+								EmptyDir: &v1.EmptyDirVolumeSource{},
+							},
+						},
+					},
+					Containers: []v1.Container{
+						{
+							Name:    "c",
+							Image:   framework.BusyBoxImage,
+							Command: []string{},
+							VolumeMounts: []v1.VolumeMount{
+								{
+									MountPath: "/data",
+									Name:      "data",
+								},
+							},
+						},
+					},
+				},
+			},
+		},
+	}
+	switch behavior {
+	case "notTerminate":
+		job.Spec.Template.Spec.Containers[0].Command = []string{"sleep", "1000000"}
+	case "fail":
+		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit 1"}
+	case "succeed":
+		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit 0"}
+	case "randomlySucceedOrFail":
+		// Bash's $RANDOM generates pseudorandom int in range 0 - 32767.
+		// Dividing by 16384 gives roughly 50/50 chance of success.
+		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit $(( $RANDOM / 16384 ))"}
+	case "failOnce":
+		// Fail the first the container of the pod is run, and
+		// succeed the second time. Checks for file on emptydir.
+		// If present, succeed.  If not, create but fail.
+		// Note that this cannot be used with RestartNever because
+		// it always fails the first time for a pod.
+		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "if [[ -r /data/foo ]] ; then exit 0 ; else touch /data/foo ; exit 1 ; fi"}
+	}
+	return job
+}
+
+// FinishTime returns finish time of the specified job.
+func FinishTime(finishedJob *batchv1.Job) metav1.Time {
+	var finishTime metav1.Time
+	for _, c := range finishedJob.Status.Conditions {
+		if (c.Type == batchv1.JobComplete || c.Type == batchv1.JobFailed) && c.Status == v1.ConditionTrue {
+			return c.LastTransitionTime
+		}
+	}
+	return finishTime
+}
diff --git a/test/e2e/framework/job/rest.go b/test/e2e/framework/job/rest.go
new file mode 100644
index 00000000000..d4b6a9e534b
--- /dev/null
+++ b/test/e2e/framework/job/rest.go
@@ -0,0 +1,82 @@
+/*
+Copyright 2019 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package job
+
+import (
+	"fmt"
+
+	batch "k8s.io/api/batch/v1"
+	"k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/wait"
+	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// GetJob uses c to get the Job in namespace ns named name. If the returned error is nil, the returned Job is valid.
+func GetJob(c clientset.Interface, ns, name string) (*batch.Job, error) {
+	return c.BatchV1().Jobs(ns).Get(name, metav1.GetOptions{})
+}
+
+// GetJobPods returns a list of Pods belonging to a Job.
+func GetJobPods(c clientset.Interface, ns, jobName string) (*v1.PodList, error) {
+	label := labels.SelectorFromSet(labels.Set(map[string]string{JobSelectorKey: jobName}))
+	options := metav1.ListOptions{LabelSelector: label.String()}
+	return c.CoreV1().Pods(ns).List(options)
+}
+
+// CreateJob uses c to create job in namespace ns. If the returned error is nil, the returned Job is valid and has
+// been created.
+func CreateJob(c clientset.Interface, ns string, job *batch.Job) (*batch.Job, error) {
+	return c.BatchV1().Jobs(ns).Create(job)
+}
+
+// UpdateJob uses c to updated job in namespace ns. If the returned error is nil, the returned Job is valid and has
+// been updated.
+func UpdateJob(c clientset.Interface, ns string, job *batch.Job) (*batch.Job, error) {
+	return c.BatchV1().Jobs(ns).Update(job)
+}
+
+// UpdateJobWithRetries updates job with retries.
+func UpdateJobWithRetries(c clientset.Interface, namespace, name string, applyUpdate func(*batch.Job)) (job *batch.Job, err error) {
+	jobs := c.BatchV1().Jobs(namespace)
+	var updateErr error
+	pollErr := wait.PollImmediate(framework.Poll, JobTimeout, func() (bool, error) {
+		if job, err = jobs.Get(name, metav1.GetOptions{}); err != nil {
+			return false, err
+		}
+		// Apply the update, then attempt to push it to the apiserver.
+		applyUpdate(job)
+		if job, err = jobs.Update(job); err == nil {
+			framework.Logf("Updating job %s", name)
+			return true, nil
+		}
+		updateErr = err
+		return false, nil
+	})
+	if pollErr == wait.ErrWaitTimeout {
+		pollErr = fmt.Errorf("couldn't apply the provided updated to job %q: %v", name, updateErr)
+	}
+	return job, pollErr
+}
+
+// DeleteJob uses c to delete the Job named name in namespace ns. If the returned error is nil, the Job has been
+// deleted.
+func DeleteJob(c clientset.Interface, ns, name string) error {
+	return c.BatchV1().Jobs(ns).Delete(name, nil)
+}
diff --git a/test/e2e/framework/job/wait.go b/test/e2e/framework/job/wait.go
new file mode 100644
index 00000000000..d74509e6cb7
--- /dev/null
+++ b/test/e2e/framework/job/wait.go
@@ -0,0 +1,142 @@
+/*
+Copyright 2019 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package job
+
+import (
+	"time"
+
+	batchv1 "k8s.io/api/batch/v1"
+	"k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/errors"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/wait"
+	clientset "k8s.io/client-go/kubernetes"
+	jobutil "k8s.io/kubernetes/pkg/controller/job"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// WaitForAllJobPodsRunning wait for all pods for the Job named JobName in namespace ns to become Running.  Only use
+// when pods will run for a long time, or it will be racy.
+func WaitForAllJobPodsRunning(c clientset.Interface, ns, jobName string, parallelism int32) error {
+	return wait.Poll(framework.Poll, JobTimeout, func() (bool, error) {
+		pods, err := GetJobPods(c, ns, jobName)
+		if err != nil {
+			return false, err
+		}
+		count := int32(0)
+		for _, p := range pods.Items {
+			if p.Status.Phase == v1.PodRunning {
+				count++
+			}
+		}
+		return count == parallelism, nil
+	})
+}
+
+// WaitForJobComplete uses c to wait for completions to complete for the Job jobName in namespace ns.
+func WaitForJobComplete(c clientset.Interface, ns, jobName string, completions int32) error {
+	return wait.Poll(framework.Poll, JobTimeout, func() (bool, error) {
+		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
+		if err != nil {
+			return false, err
+		}
+		return curr.Status.Succeeded == completions, nil
+	})
+}
+
+// WaitForJobFinish uses c to wait for the Job jobName in namespace ns to finish (either Failed or Complete).
+func WaitForJobFinish(c clientset.Interface, ns, jobName string) error {
+	return wait.PollImmediate(framework.Poll, JobTimeout, func() (bool, error) {
+		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
+		if err != nil {
+			return false, err
+		}
+		return jobutil.IsJobFinished(curr), nil
+	})
+}
+
+// WaitForJobFailure uses c to wait for up to timeout for the Job named jobName in namespace ns to fail.
+func WaitForJobFailure(c clientset.Interface, ns, jobName string, timeout time.Duration, reason string) error {
+	return wait.Poll(framework.Poll, timeout, func() (bool, error) {
+		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
+		if err != nil {
+			return false, err
+		}
+		for _, c := range curr.Status.Conditions {
+			if c.Type == batchv1.JobFailed && c.Status == v1.ConditionTrue {
+				if reason == "" || reason == c.Reason {
+					return true, nil
+				}
+			}
+		}
+		return false, nil
+	})
+}
+
+// WaitForJobGone uses c to wait for up to timeout for the Job named jobName in namespace ns to be removed.
+func WaitForJobGone(c clientset.Interface, ns, jobName string, timeout time.Duration) error {
+	return wait.Poll(framework.Poll, timeout, func() (bool, error) {
+		_, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
+		if errors.IsNotFound(err) {
+			return true, nil
+		}
+		return false, err
+	})
+}
+
+// CheckForAllJobPodsRunning uses c to check in the Job named jobName in ns is running. If the returned error is not
+// nil the returned bool is true if the Job is running.
+func CheckForAllJobPodsRunning(c clientset.Interface, ns, jobName string, parallelism int32) (bool, error) {
+	label := labels.SelectorFromSet(labels.Set(map[string]string{JobSelectorKey: jobName}))
+	options := metav1.ListOptions{LabelSelector: label.String()}
+	pods, err := c.CoreV1().Pods(ns).List(options)
+	if err != nil {
+		return false, err
+	}
+	count := int32(0)
+	for _, p := range pods.Items {
+		if p.Status.Phase == v1.PodRunning {
+			count++
+		}
+	}
+	return count == parallelism, nil
+}
+
+// WaitForAllJobPodsGone waits for all pods for the Job named jobName in namespace ns
+// to be deleted.
+func WaitForAllJobPodsGone(c clientset.Interface, ns, jobName string) error {
+	return wait.PollImmediate(framework.Poll, JobTimeout, func() (bool, error) {
+		pods, err := GetJobPods(c, ns, jobName)
+		if err != nil {
+			return false, err
+		}
+		return len(pods.Items) == 0, nil
+	})
+}
+
+// WaitForJobDeleting uses c to wait for the Job jobName in namespace ns to have
+// a non-nil deletionTimestamp (i.e. being deleted).
+func WaitForJobDeleting(c clientset.Interface, ns, jobName string) error {
+	return wait.PollImmediate(framework.Poll, JobTimeout, func() (bool, error) {
+		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
+		if err != nil {
+			return false, err
+		}
+		return curr.ObjectMeta.DeletionTimestamp != nil, nil
+	})
+}
diff --git a/test/e2e/framework/jobs_util.go b/test/e2e/framework/jobs_util.go
deleted file mode 100644
index 13fbb7db8e3..00000000000
--- a/test/e2e/framework/jobs_util.go
+++ /dev/null
@@ -1,318 +0,0 @@
-/*
-Copyright 2017 The Kubernetes Authors.
-
-Licensed under the Apache License, Version 2.0 (the "License");
-you may not use this file except in compliance with the License.
-You may obtain a copy of the License at
-
-    http://www.apache.org/licenses/LICENSE-2.0
-
-Unless required by applicable law or agreed to in writing, software
-distributed under the License is distributed on an "AS IS" BASIS,
-WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and
-limitations under the License.
-*/
-
-package framework
-
-import (
-	"fmt"
-	"time"
-
-	batch "k8s.io/api/batch/v1"
-	"k8s.io/api/core/v1"
-	"k8s.io/apimachinery/pkg/api/errors"
-	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/apimachinery/pkg/labels"
-	"k8s.io/apimachinery/pkg/util/wait"
-	clientset "k8s.io/client-go/kubernetes"
-	jobutil "k8s.io/kubernetes/pkg/controller/job"
-)
-
-const (
-	// JobTimeout is how long to wait for a job to finish.
-	JobTimeout = 15 * time.Minute
-
-	// JobSelectorKey is a job selector name
-	JobSelectorKey = "job"
-)
-
-// NewTestJob returns a Job which does one of several testing behaviors. notTerminate starts a Job that will run
-// effectively forever. fail starts a Job that will fail immediately. succeed starts a Job that will succeed
-// immediately. randomlySucceedOrFail starts a Job that will succeed or fail randomly. failOnce fails the Job the
-// first time it is run and succeeds subsequently. name is the Name of the Job. RestartPolicy indicates the restart
-// policy of the containers in which the Pod is running. Parallelism is the Job's parallelism, and completions is the
-// Job's required number of completions.
-func NewTestJob(behavior, name string, rPol v1.RestartPolicy, parallelism, completions int32, activeDeadlineSeconds *int64, backoffLimit int32) *batch.Job {
-	job := &batch.Job{
-		ObjectMeta: metav1.ObjectMeta{
-			Name: name,
-		},
-		TypeMeta: metav1.TypeMeta{
-			Kind: "Job",
-		},
-		Spec: batch.JobSpec{
-			ActiveDeadlineSeconds: activeDeadlineSeconds,
-			Parallelism:           &parallelism,
-			Completions:           &completions,
-			BackoffLimit:          &backoffLimit,
-			ManualSelector:        newBool(false),
-			Template: v1.PodTemplateSpec{
-				ObjectMeta: metav1.ObjectMeta{
-					Labels: map[string]string{JobSelectorKey: name},
-				},
-				Spec: v1.PodSpec{
-					RestartPolicy: rPol,
-					Volumes: []v1.Volume{
-						{
-							Name: "data",
-							VolumeSource: v1.VolumeSource{
-								EmptyDir: &v1.EmptyDirVolumeSource{},
-							},
-						},
-					},
-					Containers: []v1.Container{
-						{
-							Name:    "c",
-							Image:   BusyBoxImage,
-							Command: []string{},
-							VolumeMounts: []v1.VolumeMount{
-								{
-									MountPath: "/data",
-									Name:      "data",
-								},
-							},
-						},
-					},
-				},
-			},
-		},
-	}
-	switch behavior {
-	case "notTerminate":
-		job.Spec.Template.Spec.Containers[0].Command = []string{"sleep", "1000000"}
-	case "fail":
-		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit 1"}
-	case "succeed":
-		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit 0"}
-	case "randomlySucceedOrFail":
-		// Bash's $RANDOM generates pseudorandom int in range 0 - 32767.
-		// Dividing by 16384 gives roughly 50/50 chance of success.
-		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "exit $(( $RANDOM / 16384 ))"}
-	case "failOnce":
-		// Fail the first the container of the pod is run, and
-		// succeed the second time. Checks for file on emptydir.
-		// If present, succeed.  If not, create but fail.
-		// Note that this cannot be used with RestartNever because
-		// it always fails the first time for a pod.
-		job.Spec.Template.Spec.Containers[0].Command = []string{"/bin/sh", "-c", "if [[ -r /data/foo ]] ; then exit 0 ; else touch /data/foo ; exit 1 ; fi"}
-	}
-	return job
-}
-
-// GetJob uses c to get the Job in namespace ns named name. If the returned error is nil, the returned Job is valid.
-func GetJob(c clientset.Interface, ns, name string) (*batch.Job, error) {
-	return c.BatchV1().Jobs(ns).Get(name, metav1.GetOptions{})
-}
-
-// CreateJob uses c to create job in namespace ns. If the returned error is nil, the returned Job is valid and has
-// been created.
-func CreateJob(c clientset.Interface, ns string, job *batch.Job) (*batch.Job, error) {
-	return c.BatchV1().Jobs(ns).Create(job)
-}
-
-// UpdateJob uses c to updated job in namespace ns. If the returned error is nil, the returned Job is valid and has
-// been updated.
-func UpdateJob(c clientset.Interface, ns string, job *batch.Job) (*batch.Job, error) {
-	return c.BatchV1().Jobs(ns).Update(job)
-}
-
-// UpdateJobFunc updates the job object. It retries if there is a conflict, throw out error if
-// there is any other errors. name is the job name, updateFn is the function updating the
-// job object.
-func UpdateJobFunc(c clientset.Interface, ns, name string, updateFn func(job *batch.Job)) {
-	ExpectNoError(wait.Poll(time.Millisecond*500, time.Second*30, func() (bool, error) {
-		job, err := GetJob(c, ns, name)
-		if err != nil {
-			return false, fmt.Errorf("failed to get pod %q: %v", name, err)
-		}
-		updateFn(job)
-		_, err = UpdateJob(c, ns, job)
-		if err == nil {
-			Logf("Successfully updated job %q", name)
-			return true, nil
-		}
-		if errors.IsConflict(err) {
-			Logf("Conflicting update to job %q, re-get and re-update: %v", name, err)
-			return false, nil
-		}
-		return false, fmt.Errorf("failed to update job %q: %v", name, err)
-	}))
-}
-
-// DeleteJob uses c to delete the Job named name in namespace ns. If the returned error is nil, the Job has been
-// deleted.
-func DeleteJob(c clientset.Interface, ns, name string) error {
-	return c.BatchV1().Jobs(ns).Delete(name, nil)
-}
-
-// GetJobPods returns a list of Pods belonging to a Job.
-func GetJobPods(c clientset.Interface, ns, jobName string) (*v1.PodList, error) {
-	label := labels.SelectorFromSet(labels.Set(map[string]string{JobSelectorKey: jobName}))
-	options := metav1.ListOptions{LabelSelector: label.String()}
-	return c.CoreV1().Pods(ns).List(options)
-}
-
-// WaitForAllJobPodsRunning wait for all pods for the Job named JobName in namespace ns to become Running.  Only use
-// when pods will run for a long time, or it will be racy.
-func WaitForAllJobPodsRunning(c clientset.Interface, ns, jobName string, parallelism int32) error {
-	return wait.Poll(Poll, JobTimeout, func() (bool, error) {
-		pods, err := GetJobPods(c, ns, jobName)
-		if err != nil {
-			return false, err
-		}
-		count := int32(0)
-		for _, p := range pods.Items {
-			if p.Status.Phase == v1.PodRunning {
-				count++
-			}
-		}
-		return count == parallelism, nil
-	})
-}
-
-// WaitForJobComplete uses c to wait for completions to complete for the Job jobName in namespace ns.
-func WaitForJobComplete(c clientset.Interface, ns, jobName string, completions int32) error {
-	return wait.Poll(Poll, JobTimeout, func() (bool, error) {
-		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
-		if err != nil {
-			return false, err
-		}
-		return curr.Status.Succeeded == completions, nil
-	})
-}
-
-// WaitForJobFinish uses c to wait for the Job jobName in namespace ns to finish (either Failed or Complete).
-func WaitForJobFinish(c clientset.Interface, ns, jobName string) error {
-	return wait.PollImmediate(Poll, JobTimeout, func() (bool, error) {
-		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
-		if err != nil {
-			return false, err
-		}
-		return jobutil.IsJobFinished(curr), nil
-	})
-}
-
-// WaitForJobFailure uses c to wait for up to timeout for the Job named jobName in namespace ns to fail.
-func WaitForJobFailure(c clientset.Interface, ns, jobName string, timeout time.Duration, reason string) error {
-	return wait.Poll(Poll, timeout, func() (bool, error) {
-		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
-		if err != nil {
-			return false, err
-		}
-		for _, c := range curr.Status.Conditions {
-			if c.Type == batch.JobFailed && c.Status == v1.ConditionTrue {
-				if reason == "" || reason == c.Reason {
-					return true, nil
-				}
-			}
-		}
-		return false, nil
-	})
-}
-
-// WaitForJobGone uses c to wait for up to timeout for the Job named jobName in namespace ns to be removed.
-func WaitForJobGone(c clientset.Interface, ns, jobName string, timeout time.Duration) error {
-	return wait.Poll(Poll, timeout, func() (bool, error) {
-		_, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
-		if errors.IsNotFound(err) {
-			return true, nil
-		}
-		return false, err
-	})
-}
-
-// CheckForAllJobPodsRunning uses c to check in the Job named jobName in ns is running. If the returned error is not
-// nil the returned bool is true if the Job is running.
-func CheckForAllJobPodsRunning(c clientset.Interface, ns, jobName string, parallelism int32) (bool, error) {
-	label := labels.SelectorFromSet(labels.Set(map[string]string{JobSelectorKey: jobName}))
-	options := metav1.ListOptions{LabelSelector: label.String()}
-	pods, err := c.CoreV1().Pods(ns).List(options)
-	if err != nil {
-		return false, err
-	}
-	count := int32(0)
-	for _, p := range pods.Items {
-		if p.Status.Phase == v1.PodRunning {
-			count++
-		}
-	}
-	return count == parallelism, nil
-}
-
-// WaitForAllJobPodsGone waits for all pods for the Job named jobName in namespace ns
-// to be deleted.
-func WaitForAllJobPodsGone(c clientset.Interface, ns, jobName string) error {
-	return wait.PollImmediate(Poll, JobTimeout, func() (bool, error) {
-		pods, err := GetJobPods(c, ns, jobName)
-		if err != nil {
-			return false, err
-		}
-		return len(pods.Items) == 0, nil
-	})
-}
-
-func newBool(val bool) *bool {
-	p := new(bool)
-	*p = val
-	return p
-}
-
-type updateJobFunc func(*batch.Job)
-
-// UpdateJobWithRetries updates jobs with retries.
-func UpdateJobWithRetries(c clientset.Interface, namespace, name string, applyUpdate updateJobFunc) (job *batch.Job, err error) {
-	jobs := c.BatchV1().Jobs(namespace)
-	var updateErr error
-	pollErr := wait.PollImmediate(Poll, JobTimeout, func() (bool, error) {
-		if job, err = jobs.Get(name, metav1.GetOptions{}); err != nil {
-			return false, err
-		}
-		// Apply the update, then attempt to push it to the apiserver.
-		applyUpdate(job)
-		if job, err = jobs.Update(job); err == nil {
-			Logf("Updating job %s", name)
-			return true, nil
-		}
-		updateErr = err
-		return false, nil
-	})
-	if pollErr == wait.ErrWaitTimeout {
-		pollErr = fmt.Errorf("couldn't apply the provided updated to job %q: %v", name, updateErr)
-	}
-	return job, pollErr
-}
-
-// WaitForJobDeleting uses c to wait for the Job jobName in namespace ns to have
-// a non-nil deletionTimestamp (i.e. being deleted).
-func WaitForJobDeleting(c clientset.Interface, ns, jobName string) error {
-	return wait.PollImmediate(Poll, JobTimeout, func() (bool, error) {
-		curr, err := c.BatchV1().Jobs(ns).Get(jobName, metav1.GetOptions{})
-		if err != nil {
-			return false, err
-		}
-		return curr.ObjectMeta.DeletionTimestamp != nil, nil
-	})
-}
-
-// JobFinishTime returns finish time of the specified job.
-func JobFinishTime(finishedJob *batch.Job) metav1.Time {
-	var finishTime metav1.Time
-	for _, c := range finishedJob.Status.Conditions {
-		if (c.Type == batch.JobComplete || c.Type == batch.JobFailed) && c.Status == v1.ConditionTrue {
-			return c.LastTransitionTime
-		}
-	}
-	return finishTime
-}
diff --git a/test/e2e/kubectl/BUILD b/test/e2e/kubectl/BUILD
index eb665e57e7c..e42e36df51b 100644
--- a/test/e2e/kubectl/BUILD
+++ b/test/e2e/kubectl/BUILD
@@ -31,6 +31,7 @@ go_library(
         "//staging/src/k8s.io/client-go/kubernetes:go_default_library",
         "//test/e2e/common:go_default_library",
         "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/e2e/framework/testfiles:go_default_library",
         "//test/e2e/scheduling:go_default_library",
         "//test/utils:go_default_library",
diff --git a/test/e2e/kubectl/kubectl.go b/test/e2e/kubectl/kubectl.go
index 56ea3c1b278..a8910f3dc34 100644
--- a/test/e2e/kubectl/kubectl.go
+++ b/test/e2e/kubectl/kubectl.go
@@ -58,6 +58,7 @@ import (
 	"k8s.io/kubernetes/pkg/controller"
 	commonutils "k8s.io/kubernetes/test/e2e/common"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 	"k8s.io/kubernetes/test/e2e/framework/testfiles"
 	"k8s.io/kubernetes/test/e2e/scheduling"
 	testutils "k8s.io/kubernetes/test/utils"
@@ -1699,7 +1700,7 @@ metadata:
 			gomega.Expect(runOutput).To(gomega.ContainSubstring("abcd1234"))
 			gomega.Expect(runOutput).To(gomega.ContainSubstring("stdin closed"))
 
-			err := framework.WaitForJobGone(c, ns, jobName, wait.ForeverTestTimeout)
+			err := jobutil.WaitForJobGone(c, ns, jobName, wait.ForeverTestTimeout)
 			gomega.Expect(err).NotTo(gomega.HaveOccurred())
 
 			ginkgo.By("verifying the job " + jobName + " was deleted")
diff --git a/test/e2e/node/BUILD b/test/e2e/node/BUILD
index 91494701c7a..ddea86b2bde 100644
--- a/test/e2e/node/BUILD
+++ b/test/e2e/node/BUILD
@@ -36,6 +36,7 @@ go_library(
         "//staging/src/k8s.io/client-go/kubernetes:go_default_library",
         "//test/e2e/common:go_default_library",
         "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/utils:go_default_library",
         "//test/utils/image:go_default_library",
         "//vendor/github.com/onsi/ginkgo:go_default_library",
diff --git a/test/e2e/node/ttlafterfinished.go b/test/e2e/node/ttlafterfinished.go
index 849fd572634..e8634746915 100644
--- a/test/e2e/node/ttlafterfinished.go
+++ b/test/e2e/node/ttlafterfinished.go
@@ -24,6 +24,7 @@ import (
 	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/kubernetes/pkg/util/slice"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 
 	. "github.com/onsi/ginkgo"
 	. "github.com/onsi/gomega"
@@ -47,11 +48,11 @@ func cleanupJob(f *framework.Framework, job *batch.Job) {
 	removeFinalizerFunc := func(j *batch.Job) {
 		j.ObjectMeta.Finalizers = slice.RemoveString(j.ObjectMeta.Finalizers, dummyFinalizer, nil)
 	}
-	_, err := framework.UpdateJobWithRetries(c, ns, job.Name, removeFinalizerFunc)
+	_, err := jobutil.UpdateJobWithRetries(c, ns, job.Name, removeFinalizerFunc)
 	Expect(err).NotTo(HaveOccurred())
-	framework.WaitForJobGone(c, ns, job.Name, wait.ForeverTestTimeout)
+	jobutil.WaitForJobGone(c, ns, job.Name, wait.ForeverTestTimeout)
 
-	err = framework.WaitForAllJobPodsGone(c, ns, job.Name)
+	err = jobutil.WaitForAllJobPodsGone(c, ns, job.Name)
 	Expect(err).NotTo(HaveOccurred())
 }
 
@@ -64,27 +65,27 @@ func testFinishedJob(f *framework.Framework) {
 	backoffLimit := int32(2)
 	ttl := int32(10)
 
-	job := framework.NewTestJob("randomlySucceedOrFail", "rand-non-local", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
+	job := jobutil.NewTestJob("randomlySucceedOrFail", "rand-non-local", v1.RestartPolicyNever, parallelism, completions, nil, backoffLimit)
 	job.Spec.TTLSecondsAfterFinished = &ttl
 	job.ObjectMeta.Finalizers = []string{dummyFinalizer}
 	defer cleanupJob(f, job)
 
 	framework.Logf("Create a Job %s/%s with TTL", ns, job.Name)
-	job, err := framework.CreateJob(c, ns, job)
+	job, err := jobutil.CreateJob(c, ns, job)
 	Expect(err).NotTo(HaveOccurred())
 
 	framework.Logf("Wait for the Job to finish")
-	err = framework.WaitForJobFinish(c, ns, job.Name)
+	err = jobutil.WaitForJobFinish(c, ns, job.Name)
 	Expect(err).NotTo(HaveOccurred())
 
 	framework.Logf("Wait for TTL after finished controller to delete the Job")
-	err = framework.WaitForJobDeleting(c, ns, job.Name)
+	err = jobutil.WaitForJobDeleting(c, ns, job.Name)
 	Expect(err).NotTo(HaveOccurred())
 
 	framework.Logf("Check Job's deletionTimestamp and compare with the time when the Job finished")
-	job, err = framework.GetJob(c, ns, job.Name)
+	job, err = jobutil.GetJob(c, ns, job.Name)
 	Expect(err).NotTo(HaveOccurred())
-	finishTime := framework.JobFinishTime(job)
+	finishTime := jobutil.FinishTime(job)
 	finishTimeUTC := finishTime.UTC()
 	Expect(finishTime.IsZero()).NotTo(BeTrue())
 
diff --git a/test/e2e/upgrades/BUILD b/test/e2e/upgrades/BUILD
index a2465882edf..14dbcc29515 100644
--- a/test/e2e/upgrades/BUILD
+++ b/test/e2e/upgrades/BUILD
@@ -38,6 +38,7 @@ go_library(
         "//test/e2e/common:go_default_library",
         "//test/e2e/framework:go_default_library",
         "//test/e2e/framework/gpu:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/e2e/framework/testfiles:go_default_library",
         "//test/e2e/scheduling:go_default_library",
         "//test/utils/image:go_default_library",
diff --git a/test/e2e/upgrades/apps/BUILD b/test/e2e/upgrades/apps/BUILD
index 633213ffdb0..c3c25b0c628 100644
--- a/test/e2e/upgrades/apps/BUILD
+++ b/test/e2e/upgrades/apps/BUILD
@@ -27,6 +27,7 @@ go_library(
         "//staging/src/k8s.io/apimachinery/pkg/util/version:go_default_library",
         "//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
         "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/job:go_default_library",
         "//test/e2e/framework/replicaset:go_default_library",
         "//test/e2e/upgrades:go_default_library",
         "//test/utils/image:go_default_library",
diff --git a/test/e2e/upgrades/apps/job.go b/test/e2e/upgrades/apps/job.go
index 36612a09c13..15eacd96356 100644
--- a/test/e2e/upgrades/apps/job.go
+++ b/test/e2e/upgrades/apps/job.go
@@ -20,6 +20,7 @@ import (
 	batch "k8s.io/api/batch/v1"
 	"k8s.io/api/core/v1"
 	"k8s.io/kubernetes/test/e2e/framework"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 	"k8s.io/kubernetes/test/e2e/upgrades"
 
 	"github.com/onsi/ginkgo"
@@ -40,13 +41,13 @@ func (t *JobUpgradeTest) Setup(f *framework.Framework) {
 	t.namespace = f.Namespace.Name
 
 	ginkgo.By("Creating a job")
-	t.job = framework.NewTestJob("notTerminate", "foo", v1.RestartPolicyOnFailure, 2, 2, nil, 6)
-	job, err := framework.CreateJob(f.ClientSet, t.namespace, t.job)
+	t.job = jobutil.NewTestJob("notTerminate", "foo", v1.RestartPolicyOnFailure, 2, 2, nil, 6)
+	job, err := jobutil.CreateJob(f.ClientSet, t.namespace, t.job)
 	t.job = job
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 
 	ginkgo.By("Ensuring active pods == parallelism")
-	err = framework.WaitForAllJobPodsRunning(f.ClientSet, t.namespace, job.Name, 2)
+	err = jobutil.WaitForAllJobPodsRunning(f.ClientSet, t.namespace, job.Name, 2)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 }
 
@@ -54,7 +55,7 @@ func (t *JobUpgradeTest) Setup(f *framework.Framework) {
 func (t *JobUpgradeTest) Test(f *framework.Framework, done <-chan struct{}, upgrade upgrades.UpgradeType) {
 	<-done
 	ginkgo.By("Ensuring active pods == parallelism")
-	running, err := framework.CheckForAllJobPodsRunning(f.ClientSet, t.namespace, t.job.Name, 2)
+	running, err := jobutil.CheckForAllJobPodsRunning(f.ClientSet, t.namespace, t.job.Name, 2)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	gomega.Expect(running).To(gomega.BeTrue())
 }
diff --git a/test/e2e/upgrades/nvidia-gpu.go b/test/e2e/upgrades/nvidia-gpu.go
index 459de01bcf6..c7d35f1fd8e 100644
--- a/test/e2e/upgrades/nvidia-gpu.go
+++ b/test/e2e/upgrades/nvidia-gpu.go
@@ -24,6 +24,7 @@ import (
 	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/kubernetes/test/e2e/framework"
 	"k8s.io/kubernetes/test/e2e/framework/gpu"
+	jobutil "k8s.io/kubernetes/test/e2e/framework/job"
 	"k8s.io/kubernetes/test/e2e/scheduling"
 	imageutils "k8s.io/kubernetes/test/utils/image"
 
@@ -54,7 +55,7 @@ func (t *NvidiaGPUUpgradeTest) Test(f *framework.Framework, done <-chan struct{}
 	t.verifyJobPodSuccess(f)
 	if upgrade == MasterUpgrade || upgrade == ClusterUpgrade {
 		// MasterUpgrade should be totally hitless.
-		job, err := framework.GetJob(f.ClientSet, f.Namespace.Name, "cuda-add")
+		job, err := jobutil.GetJob(f.ClientSet, f.Namespace.Name, "cuda-add")
 		gomega.Expect(err).NotTo(gomega.HaveOccurred())
 		gomega.Expect(job.Status.Failed).To(gomega.BeZero(), "Job pods failed during master upgrade: %v", job.Status.Failed)
 	}
@@ -69,7 +70,7 @@ func (t *NvidiaGPUUpgradeTest) Teardown(f *framework.Framework) {
 func (t *NvidiaGPUUpgradeTest) startJob(f *framework.Framework) {
 	var activeSeconds int64 = 3600
 	// Specifies 100 completions to make sure the job life spans across the upgrade.
-	testJob := framework.NewTestJob("succeed", "cuda-add", v1.RestartPolicyAlways, 1, 100, &activeSeconds, 6)
+	testJob := jobutil.NewTestJob("succeed", "cuda-add", v1.RestartPolicyAlways, 1, 100, &activeSeconds, 6)
 	testJob.Spec.Template.Spec = v1.PodSpec{
 		RestartPolicy: v1.RestartPolicyOnFailure,
 		Containers: []v1.Container{
@@ -86,11 +87,11 @@ func (t *NvidiaGPUUpgradeTest) startJob(f *framework.Framework) {
 		},
 	}
 	ns := f.Namespace.Name
-	_, err := framework.CreateJob(f.ClientSet, ns, testJob)
+	_, err := jobutil.CreateJob(f.ClientSet, ns, testJob)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	framework.Logf("Created job %v", testJob)
 	ginkgo.By("Waiting for gpu job pod start")
-	err = framework.WaitForAllJobPodsRunning(f.ClientSet, ns, testJob.Name, 1)
+	err = jobutil.WaitForAllJobPodsRunning(f.ClientSet, ns, testJob.Name, 1)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	ginkgo.By("Done with gpu job pod start")
 }
@@ -99,9 +100,9 @@ func (t *NvidiaGPUUpgradeTest) startJob(f *framework.Framework) {
 func (t *NvidiaGPUUpgradeTest) verifyJobPodSuccess(f *framework.Framework) {
 	// Wait for client pod to complete.
 	ns := f.Namespace.Name
-	err := framework.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, "cuda-add", 1)
+	err := jobutil.WaitForAllJobPodsRunning(f.ClientSet, f.Namespace.Name, "cuda-add", 1)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
-	pods, err := framework.GetJobPods(f.ClientSet, f.Namespace.Name, "cuda-add")
+	pods, err := jobutil.GetJobPods(f.ClientSet, f.Namespace.Name, "cuda-add")
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	createdPod := pods.Items[0].Name
 	framework.Logf("Created pod %v", createdPod)