diff --git a/pkg/controller/node/node_controller.go b/pkg/controller/node/node_controller.go
index c81fd7164e0..a33e0b8b75a 100644
--- a/pkg/controller/node/node_controller.go
+++ b/pkg/controller/node/node_controller.go
@@ -309,14 +309,12 @@ func NewNodeController(
 
 	podInformer.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{
 		AddFunc: func(obj interface{}) {
-			nc.maybeDeleteTerminatingPod(obj)
 			pod := obj.(*v1.Pod)
 			if nc.taintManager != nil {
 				nc.taintManager.PodUpdated(nil, pod)
 			}
 		},
 		UpdateFunc: func(prev, obj interface{}) {
-			nc.maybeDeleteTerminatingPod(obj)
 			prevPod := prev.(*v1.Pod)
 			newPod := obj.(*v1.Pod)
 			if nc.taintManager != nil {
@@ -1196,55 +1194,3 @@ func (nc *Controller) ComputeZoneState(nodeReadyConditions []*v1.NodeCondition)
 		return notReadyNodes, stateNormal
 	}
 }
-
-// maybeDeleteTerminatingPod non-gracefully deletes pods that are terminating
-// that should not be gracefully terminated.
-func (nc *Controller) maybeDeleteTerminatingPod(obj interface{}) {
-	pod, ok := obj.(*v1.Pod)
-	if !ok {
-		tombstone, ok := obj.(cache.DeletedFinalStateUnknown)
-		if !ok {
-			glog.Errorf("Couldn't get object from tombstone %#v", obj)
-			return
-		}
-		pod, ok = tombstone.Obj.(*v1.Pod)
-		if !ok {
-			glog.Errorf("Tombstone contained object that is not a Pod %#v", obj)
-			return
-		}
-	}
-
-	// consider only terminating pods
-	if pod.DeletionTimestamp == nil {
-		return
-	}
-
-	node, err := nc.nodeLister.Get(pod.Spec.NodeName)
-	// if there is no such node, do nothing and let the podGC clean it up.
-	if apierrors.IsNotFound(err) {
-		return
-	}
-	if err != nil {
-		// this can only happen if the Store.KeyFunc has a problem creating
-		// a key for the pod. If it happens once, it will happen again so
-		// don't bother requeuing the pod.
-		utilruntime.HandleError(err)
-		return
-	}
-
-	// delete terminating pods that have been scheduled on
-	// nodes that do not support graceful termination
-	// TODO(mikedanese): this can be removed when we no longer
-	// guarantee backwards compatibility of master API to kubelets with
-	// versions less than 1.1.0
-	v, err := utilversion.ParseSemantic(node.Status.NodeInfo.KubeletVersion)
-	if err != nil {
-		glog.V(0).Infof("Couldn't parse version %q of node: %v", node.Status.NodeInfo.KubeletVersion, err)
-		utilruntime.HandleError(nc.forcefullyDeletePod(pod))
-		return
-	}
-	if v.LessThan(gracefulDeletionVersion) {
-		utilruntime.HandleError(nc.forcefullyDeletePod(pod))
-		return
-	}
-}
diff --git a/pkg/controller/node/nodecontroller_test.go b/pkg/controller/node/nodecontroller_test.go
index 710e5634a05..f242a2516b4 100644
--- a/pkg/controller/node/nodecontroller_test.go
+++ b/pkg/controller/node/nodecontroller_test.go
@@ -1726,9 +1726,6 @@ func TestMonitorNodeStatusMarkPodsNotReady(t *testing.T) {
 							CreationTimestamp: metav1.Date(2012, 1, 1, 0, 0, 0, 0, time.UTC),
 						},
 						Status: v1.NodeStatus{
-							NodeInfo: v1.NodeSystemInfo{
-								KubeletVersion: "v1.2.0",
-							},
 							Conditions: []v1.NodeCondition{
 								{
 									Type:   v1.NodeReady,
@@ -1752,9 +1749,6 @@ func TestMonitorNodeStatusMarkPodsNotReady(t *testing.T) {
 			},
 			timeToPass: 1 * time.Minute,
 			newNodeStatus: v1.NodeStatus{
-				NodeInfo: v1.NodeSystemInfo{
-					KubeletVersion: "v1.2.0",
-				},
 				Conditions: []v1.NodeCondition{
 					{
 						Type:   v1.NodeReady,
diff --git a/pkg/controller/node/util/controller_utils.go b/pkg/controller/node/util/controller_utils.go
index 6a860f4de9f..17258e774f4 100644
--- a/pkg/controller/node/util/controller_utils.go
+++ b/pkg/controller/node/util/controller_utils.go
@@ -152,13 +152,6 @@ func ForcefullyDeleteNode(kubeClient clientset.Interface, nodeName string) error
 // MarkAllPodsNotReady updates ready status of all pods running on
 // given node from master return true if success
 func MarkAllPodsNotReady(kubeClient clientset.Interface, node *v1.Node) error {
-	// Don't set pods to NotReady if the kubelet is running a version that
-	// doesn't understand how to correct readiness.
-	// TODO: Remove this check when we no longer guarantee backward compatibility
-	// with node versions < 1.2.0.
-	if NodeRunningOutdatedKubelet(node) {
-		return nil
-	}
 	nodeName := node.Name
 	glog.V(2).Infof("Update ready status of pods on node [%v]", nodeName)
 	opts := metav1.ListOptions{FieldSelector: fields.OneTermEqualSelector(api.PodHostField, nodeName).String()}
@@ -193,23 +186,6 @@ func MarkAllPodsNotReady(kubeClient clientset.Interface, node *v1.Node) error {
 	return fmt.Errorf("%v", strings.Join(errMsg, "; "))
 }
 
-// NodeRunningOutdatedKubelet returns true if the kubeletVersion reported
-// in the nodeInfo of the given node is "outdated", meaning < 1.2.0.
-// Older versions were inflexible and modifying pod.Status directly through
-// the apiserver would result in unexpected outcomes.
-func NodeRunningOutdatedKubelet(node *v1.Node) bool {
-	v, err := utilversion.ParseSemantic(node.Status.NodeInfo.KubeletVersion)
-	if err != nil {
-		glog.Errorf("couldn't parse version %q of node %v", node.Status.NodeInfo.KubeletVersion, err)
-		return true
-	}
-	if v.LessThan(podStatusReconciliationVersion) {
-		glog.Infof("Node %v running kubelet at (%v) which is less than the minimum version that allows nodecontroller to mark pods NotReady (%v).", node.Name, v, podStatusReconciliationVersion)
-		return true
-	}
-	return false
-}
-
 // NodeExistsInCloudProvider returns true if the node exists in the
 // cloud provider.
 func NodeExistsInCloudProvider(cloud cloudprovider.Interface, nodeName types.NodeName) (bool, error) {