diff --git a/pkg/cloudprovider/providers/gce/BUILD b/pkg/cloudprovider/providers/gce/BUILD
index ee79c5d6681..beee30431bb 100644
--- a/pkg/cloudprovider/providers/gce/BUILD
+++ b/pkg/cloudprovider/providers/gce/BUILD
@@ -41,7 +41,9 @@ go_library(
         "//pkg/client/clientset_generated/clientset:go_default_library",
         "//pkg/cloudprovider:go_default_library",
         "//pkg/controller:go_default_library",
+        "//pkg/master/ports:go_default_library",
         "//pkg/util/net/sets:go_default_library",
+        "//pkg/util/version:go_default_library",
         "//pkg/volume:go_default_library",
         "//vendor/cloud.google.com/go/compute/metadata:go_default_library",
         "//vendor/github.com/golang/glog:go_default_library",
@@ -70,11 +72,13 @@ go_test(
     name = "go_default_test",
     srcs = [
         "gce_disks_test.go",
+        "gce_healthchecks_test.go",
         "gce_test.go",
     ],
     library = ":go_default_library",
     tags = ["automanaged"],
     deps = [
+        "//pkg/api/v1:go_default_library",
         "//pkg/cloudprovider:go_default_library",
         "//vendor/google.golang.org/api/compute/v1:go_default_library",
         "//vendor/google.golang.org/api/googleapi:go_default_library",
diff --git a/pkg/cloudprovider/providers/gce/gce.go b/pkg/cloudprovider/providers/gce/gce.go
index 06919686b58..a620afcb32b 100644
--- a/pkg/cloudprovider/providers/gce/gce.go
+++ b/pkg/cloudprovider/providers/gce/gce.go
@@ -22,6 +22,7 @@ import (
 	"net/http"
 	"regexp"
 	"strings"
+	"sync"
 	"time"
 
 	"cloud.google.com/go/compute/metadata"
@@ -80,7 +81,7 @@ type GCECloud struct {
 	serviceBeta              *computebeta.Service
 	containerService         *container.Service
 	clientBuilder            controller.ControllerClientBuilder
-	ClusterId                ClusterId
+	ClusterID                ClusterID
 	projectID                string
 	region                   string
 	localZone                string   // The zone in which we are running
@@ -92,6 +93,11 @@ type GCECloud struct {
 	useMetadataServer        bool
 	operationPollRateLimiter flowcontrol.RateLimiter
 	manager                  ServiceManager
+	// sharedResourceLock is used to serialize GCE operations that may mutate shared state to
+	// prevent inconsistencies. For example, load balancers manipulation methods will take the
+	// lock to prevent shared resources from being prematurely deleted while the operation is
+	// in progress.
+	sharedResourceLock sync.Mutex
 }
 
 type ServiceManager interface {
@@ -270,10 +276,10 @@ func CreateGCECloud(projectID, region, zone string, managedZones []string, netwo
 }
 
 // Initialize takes in a clientBuilder and spawns a goroutine for watching the clusterid configmap.
-// This must be called before utilizing the funcs of gce.ClusterId
+// This must be called before utilizing the funcs of gce.ClusterID
 func (gce *GCECloud) Initialize(clientBuilder controller.ControllerClientBuilder) {
 	gce.clientBuilder = clientBuilder
-	go gce.watchClusterId()
+	go gce.watchClusterID()
 }
 
 // LoadBalancer returns an implementation of LoadBalancer for Google Compute Engine.
diff --git a/pkg/cloudprovider/providers/gce/gce_clusterid.go b/pkg/cloudprovider/providers/gce/gce_clusterid.go
index 21e533cb17b..621e927670a 100644
--- a/pkg/cloudprovider/providers/gce/gce_clusterid.go
+++ b/pkg/cloudprovider/providers/gce/gce_clusterid.go
@@ -1,5 +1,5 @@
 /*
-Copyright 2014 The Kubernetes Authors.
+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.
@@ -49,18 +49,18 @@ const (
 	updateFuncFrequency = 10 * time.Minute
 )
 
-type ClusterId struct {
+type ClusterID struct {
 	idLock     sync.RWMutex
 	client     clientset.Interface
 	cfgMapKey  string
 	store      cache.Store
-	providerId *string
-	clusterId  *string
+	providerID *string
+	clusterID  *string
 }
 
-// Continually watches for changes to the cluser id config map
-func (gce *GCECloud) watchClusterId() {
-	gce.ClusterId = ClusterId{
+// Continually watches for changes to the cluster id config map
+func (gce *GCECloud) watchClusterID() {
+	gce.ClusterID = ClusterID{
 		cfgMapKey: fmt.Sprintf("%v/%v", UIDNamespace, UIDConfigMapName),
 		client:    gce.clientBuilder.ClientOrDie("cloud-provider"),
 	}
@@ -77,8 +77,8 @@ func (gce *GCECloud) watchClusterId() {
 				return
 			}
 
-			glog.V(4).Infof("Observed new configmap for clusterid: %v, %v; setting local values", m.Name, m.Data)
-			gce.ClusterId.setIds(m)
+			glog.V(4).Infof("Observed new configmap for clusteriD: %v, %v; setting local values", m.Name, m.Data)
+			gce.ClusterID.update(m)
 		},
 		UpdateFunc: func(old, cur interface{}) {
 			m, ok := cur.(*v1.ConfigMap)
@@ -96,71 +96,71 @@ func (gce *GCECloud) watchClusterId() {
 				return
 			}
 
-			glog.V(4).Infof("Observed updated configmap for clusterid %v, %v; setting local values", m.Name, m.Data)
-			gce.ClusterId.setIds(m)
+			glog.V(4).Infof("Observed updated configmap for clusteriD %v, %v; setting local values", m.Name, m.Data)
+			gce.ClusterID.update(m)
 		},
 	}
 
-	listerWatcher := cache.NewListWatchFromClient(gce.ClusterId.client.Core().RESTClient(), "configmaps", UIDNamespace, fields.Everything())
+	listerWatcher := cache.NewListWatchFromClient(gce.ClusterID.client.Core().RESTClient(), "configmaps", UIDNamespace, fields.Everything())
 	var controller cache.Controller
-	gce.ClusterId.store, controller = cache.NewInformer(newSingleObjectListerWatcher(listerWatcher, UIDConfigMapName), &v1.ConfigMap{}, updateFuncFrequency, mapEventHandler)
+	gce.ClusterID.store, controller = cache.NewInformer(newSingleObjectListerWatcher(listerWatcher, UIDConfigMapName), &v1.ConfigMap{}, updateFuncFrequency, mapEventHandler)
 
 	controller.Run(nil)
 }
 
-// GetId returns the id which is unique to this cluster
+// GetID returns the id which is unique to this cluster
 // if federated, return the provider id (unique to the cluster)
 // if not federated, return the cluster id
-func (ci *ClusterId) GetId() (string, error) {
+func (ci *ClusterID) GetID() (string, error) {
 	if err := ci.getOrInitialize(); err != nil {
 		return "", err
 	}
 
 	ci.idLock.RLock()
 	defer ci.idLock.RUnlock()
-	if ci.clusterId == nil {
+	if ci.clusterID == nil {
 		return "", errors.New("Could not retrieve cluster id")
 	}
 
 	// If provider ID is set, (Federation is enabled) use this field
-	if ci.providerId != nil && *ci.providerId != *ci.clusterId {
-		return *ci.providerId, nil
+	if ci.providerID != nil {
+		return *ci.providerID, nil
 	}
 
-	// providerId is not set, use the cluster id
-	return *ci.clusterId, nil
+	// providerID is not set, use the cluster id
+	return *ci.clusterID, nil
 }
 
 // GetFederationId returns the id which could represent the entire Federation
 // or just the cluster if not federated.
-func (ci *ClusterId) GetFederationId() (string, bool, error) {
+func (ci *ClusterID) GetFederationId() (string, bool, error) {
 	if err := ci.getOrInitialize(); err != nil {
 		return "", false, err
 	}
 
 	ci.idLock.RLock()
 	defer ci.idLock.RUnlock()
-	if ci.clusterId == nil {
+	if ci.clusterID == nil {
 		return "", false, errors.New("Could not retrieve cluster id")
 	}
 
 	// If provider ID is not set, return false
-	if ci.providerId == nil || *ci.clusterId == *ci.providerId {
+	if ci.providerID == nil || *ci.clusterID == *ci.providerID {
 		return "", false, nil
 	}
 
-	return *ci.clusterId, true, nil
+	return *ci.clusterID, true, nil
 }
 
 // getOrInitialize either grabs the configmaps current value or defines the value
-// and sets the configmap. This is for the case of the user calling GetClusterId()
+// and sets the configmap. This is for the case of the user calling GetClusterID()
 // before the watch has begun.
-func (ci *ClusterId) getOrInitialize() error {
+func (ci *ClusterID) getOrInitialize() error {
 	if ci.store == nil {
-		return errors.New("GCECloud.ClusterId is not ready. Call Initialize() before using.")
+		return errors.New("GCECloud.ClusterID is not ready. Call Initialize() before using.")
 	}
 
-	if ci.clusterId != nil {
+	if ci.clusterID != nil {
 		return nil
 	}
 
@@ -177,7 +177,7 @@ func (ci *ClusterId) getOrInitialize() error {
 		return err
 	}
 
-	glog.V(4).Infof("Creating clusterid: %v", newId)
+	glog.V(4).Infof("Creating clusteriD: %v", newId)
 	cfg := &v1.ConfigMap{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      UIDConfigMapName,
@@ -194,12 +194,12 @@ func (ci *ClusterId) getOrInitialize() error {
 		return err
 	}
 
-	glog.V(2).Infof("Created a config map containing clusterid: %v", newId)
-	ci.setIds(cfg)
+	glog.V(2).Infof("Created a config map containing clusteriD: %v", newId)
+	ci.update(cfg)
 	return nil
 }
 
-func (ci *ClusterId) getConfigMap() (bool, error) {
+func (ci *ClusterID) getConfigMap() (bool, error) {
 	item, exists, err := ci.store.GetByKey(ci.cfgMapKey)
 	if err != nil {
 		return false, err
@@ -214,18 +214,18 @@ func (ci *ClusterId) getConfigMap() (bool, error) {
 		glog.Error(err)
 		return false, err
 	}
-	ci.setIds(m)
+	ci.update(m)
 	return true, nil
 }
 
-func (ci *ClusterId) setIds(m *v1.ConfigMap) {
+func (ci *ClusterID) update(m *v1.ConfigMap) {
 	ci.idLock.Lock()
 	defer ci.idLock.Unlock()
-	if clusterId, exists := m.Data[UIDCluster]; exists {
-		ci.clusterId = &clusterId
+	if clusterID, exists := m.Data[UIDCluster]; exists {
+		ci.clusterID = &clusterID
 	}
 	if provId, exists := m.Data[UIDProvider]; exists {
-		ci.providerId = &provId
+		ci.providerID = &provId
 	}
 }
 
diff --git a/pkg/cloudprovider/providers/gce/gce_healthchecks.go b/pkg/cloudprovider/providers/gce/gce_healthchecks.go
index bf168a029f5..6b0f6795c17 100644
--- a/pkg/cloudprovider/providers/gce/gce_healthchecks.go
+++ b/pkg/cloudprovider/providers/gce/gce_healthchecks.go
@@ -17,11 +17,23 @@ limitations under the License.
 package gce
 
 import (
+	"fmt"
 	"time"
 
+	"k8s.io/kubernetes/pkg/api/v1"
+	"k8s.io/kubernetes/pkg/master/ports"
+	utilversion "k8s.io/kubernetes/pkg/util/version"
+
+	"github.com/golang/glog"
 	compute "google.golang.org/api/compute/v1"
 )
 
+const (
+	minNodesHealthCheckVersion = "1.7.0"
+	nodesHealthCheckPath       = "/healthz"
+	lbNodesHealthCheckPort     = ports.ProxyHealthzPort
+)
+
 func newHealthcheckMetricContext(request string) *metricContext {
 	return &metricContext{
 		start:      time.Now(),
@@ -178,3 +190,59 @@ func (gce *GCECloud) ListHealthChecks() (*compute.HealthCheckList, error) {
 	v, err := gce.service.HealthChecks.List(gce.projectID).Do()
 	return v, mc.Observe(err)
 }
+
+// GetNodesHealthCheckPort returns the health check port used by the GCE load
+// balancers (l4) for performing health checks on nodes.
+func GetNodesHealthCheckPort() int32 {
+	return lbNodesHealthCheckPort
+}
+
+// getNodesHealthCheckPath returns the health check path used by the GCE load
+// balancers (l4) for performing health checks on nodes.
+func getNodesHealthCheckPath() string {
+	return nodesHealthCheckPath
+}
+
+// makeNodesHealthCheckName returns name of the health check resource used by
+// the GCE load balancers (l4) for performing health checks on nodes.
+func makeNodesHealthCheckName(clusterID string) string {
+	return fmt.Sprintf("k8s-%v-node", clusterID)
+}
+
+// MakeHealthCheckFirewallName returns the firewall name used by the GCE load
+// balancers (l4) for performing health checks.
+func MakeHealthCheckFirewallName(clusterID, hcName string, isNodesHealthCheck bool) string {
+	// TODO: Change below fwName to match the proposed schema: k8s-{clusteriD}-{namespace}-{name}-{shortid}-hc.
+	fwName := "k8s-" + hcName + "-http-hc"
+	if isNodesHealthCheck {
+		fwName = makeNodesHealthCheckName(clusterID) + "-http-hc"
+	}
+	return fwName
+}
+
+// isAtLeastMinNodesHealthCheckVersion checks if a version is higher than
+// `minNodesHealthCheckVersion`.
+func isAtLeastMinNodesHealthCheckVersion(vstring string) bool {
+	minVersion, err := utilversion.ParseGeneric(minNodesHealthCheckVersion)
+	if err != nil {
+		glog.Errorf("MinNodesHealthCheckVersion (%s) is not a valid version string: %v", minNodesHealthCheckVersion, err)
+		return false
+	}
+	version, err := utilversion.ParseGeneric(vstring)
+	if err != nil {
+		glog.Errorf("vstring (%s) is not a valid version string: %v", vstring, err)
+		return false
+	}
+	return version.AtLeast(minVersion)
+}
+
+// supportsNodesHealthCheck returns false if anyone of the nodes has version
+// lower than `minNodesHealthCheckVersion`.
+func supportsNodesHealthCheck(nodes []*v1.Node) bool {
+	for _, node := range nodes {
+		if !isAtLeastMinNodesHealthCheckVersion(node.Status.NodeInfo.KubeProxyVersion) {
+			return false
+		}
+	}
+	return true
+}
diff --git a/pkg/cloudprovider/providers/gce/gce_healthchecks_test.go b/pkg/cloudprovider/providers/gce/gce_healthchecks_test.go
new file mode 100644
index 00000000000..11be6795430
--- /dev/null
+++ b/pkg/cloudprovider/providers/gce/gce_healthchecks_test.go
@@ -0,0 +1,123 @@
+/*
+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 gce
+
+import (
+	"testing"
+
+	"k8s.io/kubernetes/pkg/api/v1"
+)
+
+func TestIsAtLeastMinNodesHealthCheckVersion(t *testing.T) {
+	testCases := []struct {
+		version string
+		expect  bool
+	}{
+		{"v1.7.1", true},
+		{"v1.7.0-alpha.2.597+276d289b90d322", true},
+		{"v1.6.0-beta.3.472+831q821c907t31a", false},
+		{"v1.5.2", false},
+	}
+
+	for _, tc := range testCases {
+		if res := isAtLeastMinNodesHealthCheckVersion(tc.version); res != tc.expect {
+			t.Errorf("%v: want %v, got %v", tc.version, tc.expect, res)
+		}
+	}
+}
+
+func TestSupportsNodesHealthCheck(t *testing.T) {
+	testCases := []struct {
+		desc   string
+		nodes  []*v1.Node
+		expect bool
+	}{
+		{
+			"All nodes support nodes health check",
+			[]*v1.Node{
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.7.1",
+						},
+					},
+				},
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.7.0-alpha.2.597+276d289b90d322",
+						},
+					},
+				},
+			},
+			true,
+		},
+		{
+			"All nodes don't support nodes health check",
+			[]*v1.Node{
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.6.0-beta.3.472+831q821c907t31a",
+						},
+					},
+				},
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.5.2",
+						},
+					},
+				},
+			},
+			false,
+		},
+		{
+			"One node doesn't support nodes health check",
+			[]*v1.Node{
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.7.1",
+						},
+					},
+				},
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.7.0-alpha.2.597+276d289b90d322",
+						},
+					},
+				},
+				{
+					Status: v1.NodeStatus{
+						NodeInfo: v1.NodeSystemInfo{
+							KubeProxyVersion: "v1.5.2",
+						},
+					},
+				},
+			},
+			false,
+		},
+	}
+
+	for _, tc := range testCases {
+		if res := supportsNodesHealthCheck(tc.nodes); res != tc.expect {
+			t.Errorf("%v: want %v, got %v", tc.desc, tc.expect, res)
+		}
+	}
+}
diff --git a/pkg/cloudprovider/providers/gce/gce_loadbalancer.go b/pkg/cloudprovider/providers/gce/gce_loadbalancer.go
index f6aa1cad3f5..dd6a09ef4d5 100644
--- a/pkg/cloudprovider/providers/gce/gce_loadbalancer.go
+++ b/pkg/cloudprovider/providers/gce/gce_loadbalancer.go
@@ -21,7 +21,6 @@ import (
 	"fmt"
 	"net"
 	"net/http"
-	"sort"
 	"strconv"
 	"strings"
 	"time"
@@ -140,6 +139,7 @@ func (gce *GCECloud) EnsureLoadBalancer(clusterName string, apiService *v1.Servi
 	}
 
 	hostNames := nodeNames(nodes)
+	supportsNodesHealthCheck := supportsNodesHealthCheck(nodes)
 	hosts, err := gce.getInstancesByNames(hostNames)
 	if err != nil {
 		return nil, err
@@ -289,13 +289,13 @@ func (gce *GCECloud) EnsureLoadBalancer(clusterName string, apiService *v1.Servi
 		// without needing to be deleted and recreated.
 		if firewallExists {
 			glog.Infof("EnsureLoadBalancer(%v(%v)): updating firewall", loadBalancerName, serviceName)
-			if err := gce.updateFirewall(loadBalancerName, gce.region, desc, sourceRanges, ports, hosts); err != nil {
+			if err := gce.updateFirewall(makeFirewallName(loadBalancerName), gce.region, desc, sourceRanges, ports, hosts); err != nil {
 				return nil, err
 			}
 			glog.Infof("EnsureLoadBalancer(%v(%v)): updated firewall", loadBalancerName, serviceName)
 		} else {
 			glog.Infof("EnsureLoadBalancer(%v(%v)): creating firewall", loadBalancerName, serviceName)
-			if err := gce.createFirewall(loadBalancerName, gce.region, desc, sourceRanges, ports, hosts); err != nil {
+			if err := gce.createFirewall(makeFirewallName(loadBalancerName), gce.region, desc, sourceRanges, ports, hosts); err != nil {
 				return nil, err
 			}
 			glog.Infof("EnsureLoadBalancer(%v(%v)): created firewall", loadBalancerName, serviceName)
@@ -310,34 +310,43 @@ func (gce *GCECloud) EnsureLoadBalancer(clusterName string, apiService *v1.Servi
 		glog.Infof("Target pool %v for Service %v/%v doesn't exist", loadBalancerName, apiService.Namespace, apiService.Name)
 	}
 
-	// Ensure health checks are created for this target pool to pass to createTargetPool for health check links
-	// Alternately, if the service has ExternalTrafficPolicy field set from Local to Global, we need to recreate
-	// the target pool without health checks. This needs to be prior to the forwarding rule deletion below otherwise
-	// it is not possible to delete just the target pool or http health checks later.
-	var hcToCreate *compute.HttpHealthCheck
-	hcExisting, err := gce.GetHttpHealthCheck(loadBalancerName)
+	clusterID, err := gce.ClusterID.GetID()
+	if err != nil {
+		return nil, fmt.Errorf("error getting cluster ID %s: %v", loadBalancerName, err)
+	}
+	// Check which health check needs to create and which health check needs to delete.
+	// Health check management is coupled with target pool operation to prevent leaking.
+	var hcToCreate, hcToDelete *compute.HttpHealthCheck
+	hcLocalTrafficExisting, err := gce.GetHttpHealthCheck(loadBalancerName)
 	if err != nil && !isHTTPErrorCode(err, http.StatusNotFound) {
-		return nil, fmt.Errorf("Error checking HTTP health check %s: %v", loadBalancerName, err)
+		return nil, fmt.Errorf("error checking HTTP health check %s: %v", loadBalancerName, err)
 	}
 	if path, healthCheckNodePort := apiservice.GetServiceHealthCheckPathPort(apiService); path != "" {
-		glog.V(4).Infof("service %v (%v) needs health checks on :%d%s)", apiService.Name, loadBalancerName, healthCheckNodePort, path)
-		if err != nil {
-			// This logic exists to detect a transition for a pre-existing service and turn on
-			// the tpNeedsUpdate flag to delete/recreate fwdrule/tpool adding the health check
-			// to the target pool.
-			glog.V(2).Infof("ExternalTrafficPolicy field set to Local on new or pre-existing service")
+		glog.V(4).Infof("service %v (%v) needs local traffic health checks on: %d%s)", apiService.Name, loadBalancerName, healthCheckNodePort, path)
+		if hcLocalTrafficExisting == nil {
+			// This logic exists to detect a transition for non-OnlyLocal to OnlyLocal service
+			// turn on the tpNeedsUpdate flag to delete/recreate fwdrule/tpool updating the
+			// target pool to use local traffic health check.
+			glog.V(2).Infof("Updating from nodes health checks to local traffic health checks for service %v LB %v", apiService.Name, loadBalancerName)
+			if supportsNodesHealthCheck {
+				hcToDelete = makeHttpHealthCheck(makeNodesHealthCheckName(clusterID), getNodesHealthCheckPath(), GetNodesHealthCheckPort())
+			}
 			tpNeedsUpdate = true
 		}
-		hcToCreate, err = gce.ensureHttpHealthCheck(loadBalancerName, path, healthCheckNodePort)
-		if err != nil {
-			return nil, fmt.Errorf("Failed to ensure health check for localized service %v on node port %v: %v", loadBalancerName, healthCheckNodePort, err)
-		}
+		hcToCreate = makeHttpHealthCheck(loadBalancerName, path, healthCheckNodePort)
 	} else {
-		glog.V(4).Infof("service %v does not need health checks", apiService.Name)
-		if err == nil {
-			glog.V(2).Infof("Deleting stale health checks for service %v LB %v", apiService.Name, loadBalancerName)
+		glog.V(4).Infof("Service %v needs nodes health checks.", apiService.Name)
+		if hcLocalTrafficExisting != nil {
+			// This logic exists to detect a transition from OnlyLocal to non-OnlyLocal service
+			// and turn on the tpNeedsUpdate flag to delete/recreate fwdrule/tpool updating the
+			// target pool to use nodes health check.
+			glog.V(2).Infof("Updating from local traffic health checks to nodes health checks for service %v LB %v", apiService.Name, loadBalancerName)
+			hcToDelete = hcLocalTrafficExisting
 			tpNeedsUpdate = true
 		}
+		if supportsNodesHealthCheck {
+			hcToCreate = makeHttpHealthCheck(makeNodesHealthCheckName(clusterID), getNodesHealthCheckPath(), GetNodesHealthCheckPort())
+		}
 	}
 	// Now we get to some slightly more interesting logic.
 	// First, neither target pools nor forwarding rules can be updated in place -
@@ -357,17 +366,12 @@ func (gce *GCECloud) EnsureLoadBalancer(clusterName string, apiService *v1.Servi
 		glog.Infof("EnsureLoadBalancer(%v(%v)): deleted forwarding rule", loadBalancerName, serviceName)
 	}
 	if tpExists && tpNeedsUpdate {
-		// Generate the list of health checks for this target pool to pass to deleteTargetPool
-		if path, _ := apiservice.GetServiceHealthCheckPathPort(apiService); path != "" {
-			var err error
-			hcExisting, err = gce.GetHttpHealthCheck(loadBalancerName)
-			if err != nil && !isHTTPErrorCode(err, http.StatusNotFound) {
-				glog.Infof("Failed to retrieve health check %v:%v", loadBalancerName, err)
-			}
+		// Pass healthchecks to deleteTargetPool to cleanup health checks after cleaning up the target pool itself.
+		var hcNames []string
+		if hcToDelete != nil {
+			hcNames = append(hcNames, hcToDelete.Name)
 		}
-
-		// Pass healthchecks to deleteTargetPool to cleanup health checks prior to cleaning up the target pool itself.
-		if err := gce.deleteTargetPool(loadBalancerName, gce.region, hcExisting); err != nil {
+		if err := gce.deleteTargetPool(loadBalancerName, gce.region, hcNames...); err != nil {
 			return nil, fmt.Errorf("failed to delete existing target pool %s for load balancer update: %v", loadBalancerName, err)
 		}
 		glog.Infof("EnsureLoadBalancer(%v(%v)): deleted target pool", loadBalancerName, serviceName)
@@ -381,11 +385,11 @@ func (gce *GCECloud) EnsureLoadBalancer(clusterName string, apiService *v1.Servi
 			createInstances = createInstances[:maxTargetPoolCreateInstances]
 		}
 		// Pass healthchecks to createTargetPool which needs them as health check links in the target pool
-		if err := gce.createTargetPool(loadBalancerName, serviceName.String(), gce.region, createInstances, affinityType, hcToCreate); err != nil {
+		if err := gce.createTargetPool(loadBalancerName, serviceName.String(), ipAddress, gce.region, createInstances, affinityType, hcToCreate); err != nil {
 			return nil, fmt.Errorf("failed to create target pool %s: %v", loadBalancerName, err)
 		}
 		if hcToCreate != nil {
-			glog.Infof("EnsureLoadBalancer(%v(%v)): created health checks for target pool", loadBalancerName, serviceName)
+			glog.Infof("EnsureLoadBalancer(%v(%v)): created health checks %v for target pool", loadBalancerName, serviceName, hcToCreate.Name)
 		}
 		if len(hosts) <= maxTargetPoolCreateInstances {
 			glog.Infof("EnsureLoadBalancer(%v(%v)): created target pool", loadBalancerName, serviceName)
@@ -447,18 +451,29 @@ func (gce *GCECloud) EnsureLoadBalancerDeleted(clusterName string, service *v1.S
 	glog.V(2).Infof("EnsureLoadBalancerDeleted(%v, %v, %v, %v, %v)", clusterName, service.Namespace, service.Name, loadBalancerName,
 		gce.region)
 
-	var hc *compute.HttpHealthCheck
+	var hcNames []string
 	if path, _ := apiservice.GetServiceHealthCheckPathPort(service); path != "" {
-		var err error
-		hc, err = gce.GetHttpHealthCheck(loadBalancerName)
+		hcToDelete, err := gce.GetHttpHealthCheck(loadBalancerName)
 		if err != nil && !isHTTPErrorCode(err, http.StatusNotFound) {
 			glog.Infof("Failed to retrieve health check %v:%v", loadBalancerName, err)
 			return err
 		}
+		hcNames = append(hcNames, hcToDelete.Name)
+	} else {
+		clusterID, err := gce.ClusterID.GetID()
+		if err != nil {
+			return fmt.Errorf("error getting cluster ID %s: %v", loadBalancerName, err)
+		}
+		// EnsureLoadBalancerDeleted() could be triggered by changing service from
+		// LoadBalancer type to others. In this case we have no idea whether it was
+		// using local traffic health check or nodes health check. Attempt to delete
+		// both to prevent leaking.
+		hcNames = append(hcNames, loadBalancerName)
+		hcNames = append(hcNames, makeNodesHealthCheckName(clusterID))
 	}
 
 	errs := utilerrors.AggregateGoroutines(
-		func() error { return gce.deleteFirewall(loadBalancerName, gce.region) },
+		func() error { return gce.deleteFirewall(makeFirewallName(loadBalancerName), gce.region) },
 		// Even though we don't hold on to static IPs for load balancers, it's
 		// possible that EnsureLoadBalancer left one around in a failed
 		// creation/update attempt, so make sure we clean it up here just in case.
@@ -469,7 +484,7 @@ func (gce *GCECloud) EnsureLoadBalancerDeleted(clusterName string, service *v1.S
 			if err := gce.deleteForwardingRule(loadBalancerName, gce.region); err != nil {
 				return err
 			}
-			if err := gce.deleteTargetPool(loadBalancerName, gce.region, hc); err != nil {
+			if err := gce.deleteTargetPool(loadBalancerName, gce.region, hcNames...); err != nil {
 				return err
 			}
 			return nil
@@ -509,15 +524,15 @@ func (gce *GCECloud) deleteForwardingRule(name, region string) error {
 }
 
 // DeleteTargetPool deletes the given target pool.
-func (gce *GCECloud) DeleteTargetPool(name string, hc *compute.HttpHealthCheck) error {
+func (gce *GCECloud) DeleteTargetPool(name string, hcNames ...string) error {
 	region, err := GetGCERegion(gce.localZone)
 	if err != nil {
 		return err
 	}
-	return gce.deleteTargetPool(name, region, hc)
+	return gce.deleteTargetPool(name, region, hcNames...)
 }
 
-func (gce *GCECloud) deleteTargetPool(name, region string, hc *compute.HttpHealthCheck) error {
+func (gce *GCECloud) deleteTargetPool(name, region string, hcNames ...string) error {
 	mc := newTargetPoolMetricContext("delete", region)
 	op, err := gce.service.TargetPools.Delete(gce.projectID, region, name).Do()
 
@@ -535,41 +550,86 @@ func (gce *GCECloud) deleteTargetPool(name, region string, hc *compute.HttpHealt
 	}
 
 	// Deletion of health checks is allowed only after the TargetPool reference is deleted
-	if hc != nil {
-		glog.Infof("Deleting health check %v", hc.Name)
-		if err := gce.DeleteHttpHealthCheck(hc.Name); err != nil {
-			glog.Warningf("Failed to delete health check %v: %v", hc, err)
+	for _, hcName := range hcNames {
+		if err = func() error {
+			// Check whether it is nodes health check, which has different name from the load-balancer.
+			isNodesHealthCheck := hcName != name
+			if isNodesHealthCheck {
+				// Lock to prevent deleting necessary nodes health check before it gets attached
+				// to target pool.
+				gce.sharedResourceLock.Lock()
+				defer gce.sharedResourceLock.Unlock()
+			}
+			glog.Infof("Deleting health check %v", hcName)
+			if err := gce.DeleteHttpHealthCheck(hcName); err != nil {
+				// Delete nodes health checks will fail if any other target pool is using it.
+				if isInUsedByError(err) {
+					glog.V(4).Infof("Health check %v is in used: %v.", hcName, err)
+					return nil
+				} else if !isHTTPErrorCode(err, http.StatusNotFound) {
+					glog.Warningf("Failed to delete health check %v: %v", hcName, err)
+					return err
+				}
+				// StatusNotFound could happen when:
+				// - This is the first attempt but we pass in a healthcheck that is already deleted
+				//   to prevent leaking.
+				// - This is the first attempt but user manually deleted the heathcheck.
+				// - This is a retry and in previous round we failed to delete the healthcheck firewall
+				//   after deleted the healthcheck.
+				// We continue to delete the healthcheck firewall to prevent leaking.
+				glog.V(4).Infof("Health check %v is already deleted.", hcName)
+			}
+			clusterID, err := gce.ClusterID.GetID()
+			if err != nil {
+				return fmt.Errorf("error getting cluster ID: %v", err)
+			}
+			// If health check is deleted without error, it means no load-balancer is using it.
+			// So we should delete the health check firewall as well.
+			fwName := MakeHealthCheckFirewallName(clusterID, hcName, isNodesHealthCheck)
+			glog.Infof("Deleting firewall %v.", fwName)
+			if err := gce.DeleteFirewall(fwName); err != nil {
+				if isHTTPErrorCode(err, http.StatusNotFound) {
+					glog.V(4).Infof("Firewall %v is already deleted.", fwName)
+					return nil
+				}
+				return err
+			}
+			return nil
+		}(); err != nil {
 			return err
 		}
-	} else {
-		// This is a HC cleanup attempt to prevent stale HCs when errors are encountered
-		// during HC deletion in a prior pass through EnsureLoadBalancer.
-		// The HC name matches the load balancer name - normally this is expected to fail.
-		if err := gce.DeleteHttpHealthCheck(name); err == nil {
-			// We only print a warning if this deletion actually succeeded (which
-			// means there was indeed a stale health check with the LB name.
-			glog.Warningf("Deleted stale http health check for LB: %s", name)
-		}
 	}
+
 	return nil
 }
 
-func (gce *GCECloud) createTargetPool(name, serviceName, region string, hosts []*gceInstance, affinityType v1.ServiceAffinity, hc *compute.HttpHealthCheck) error {
-	var instances []string
-	for _, host := range hosts {
-		instances = append(instances, makeHostURL(gce.projectID, host.Zone, host.Name))
-	}
+func (gce *GCECloud) createTargetPool(name, serviceName, ipAddress, region string, hosts []*gceInstance, affinityType v1.ServiceAffinity, hc *compute.HttpHealthCheck) error {
 	// health check management is coupled with targetPools to prevent leaks. A
 	// target pool is the only thing that requires a health check, so we delete
 	// associated checks on teardown, and ensure checks on setup.
 	hcLinks := []string{}
 	if hc != nil {
+		// Check whether it is nodes health check, which has different name from the load-balancer.
+		isNodesHealthCheck := hc.Name != name
+		if isNodesHealthCheck {
+			// Lock to prevent necessary nodes health check / firewall gets deleted.
+			gce.sharedResourceLock.Lock()
+			defer gce.sharedResourceLock.Unlock()
+		}
+		if err := gce.ensureHttpHealthCheckFirewall(serviceName, ipAddress, gce.region, hosts, hc.Name, int32(hc.Port), isNodesHealthCheck); err != nil {
+			return err
+		}
 		var err error
-		if hc, err = gce.ensureHttpHealthCheck(name, hc.RequestPath, int32(hc.Port)); err != nil || hc == nil {
+		if hc, err = gce.ensureHttpHealthCheck(hc.Name, hc.RequestPath, int32(hc.Port)); err != nil || hc == nil {
 			return fmt.Errorf("Failed to ensure health check for %v port %d path %v: %v", name, hc.Port, hc.RequestPath, err)
 		}
 		hcLinks = append(hcLinks, hc.SelfLink)
 	}
+
+	var instances []string
+	for _, host := range hosts {
+		instances = append(instances, makeHostURL(gce.projectID, host.Zone, host.Name))
+	}
 	glog.Infof("Creating targetpool %v with %d healthchecks", name, len(hcLinks))
 	pool := &compute.TargetPool{
 		Name:            name,
@@ -651,8 +711,8 @@ func (gce *GCECloud) targetPoolURL(name, region string) string {
 	return fmt.Sprintf("https://www.googleapis.com/compute/v1/projects/%s/regions/%s/targetPools/%s", gce.projectID, region, name)
 }
 
-func (gce *GCECloud) ensureHttpHealthCheck(name, path string, port int32) (hc *compute.HttpHealthCheck, err error) {
-	newHC := &compute.HttpHealthCheck{
+func makeHttpHealthCheck(name, path string, port int32) *compute.HttpHealthCheck {
+	return &compute.HttpHealthCheck{
 		Name:               name,
 		Port:               int64(port),
 		RequestPath:        path,
@@ -663,7 +723,10 @@ func (gce *GCECloud) ensureHttpHealthCheck(name, path string, port int32) (hc *c
 		HealthyThreshold:   gceHcHealthyThreshold,
 		UnhealthyThreshold: gceHcUnhealthyThreshold,
 	}
+}
 
+func (gce *GCECloud) ensureHttpHealthCheck(name, path string, port int32) (hc *compute.HttpHealthCheck, err error) {
+	newHC := makeHttpHealthCheck(name, path, port)
 	hc, err = gce.GetHttpHealthCheck(name)
 	if hc == nil || err != nil && isHTTPErrorCode(err, http.StatusNotFound) {
 		glog.Infof("Did not find health check %v, creating port %v path %v", name, port, path)
@@ -843,7 +906,7 @@ func (gce *GCECloud) firewallNeedsUpdate(name, serviceName, region, ipAddress st
 		if isHTTPErrorCode(err, http.StatusNotFound) {
 			return false, true, nil
 		}
-		return false, false, fmt.Errorf("error getting load balancer's target pool: %v", err)
+		return false, false, fmt.Errorf("error getting load balancer's firewall: %v", err)
 	}
 	if fw.Description != makeFirewallDescription(serviceName, ipAddress) {
 		return true, true, nil
@@ -856,7 +919,7 @@ func (gce *GCECloud) firewallNeedsUpdate(name, serviceName, region, ipAddress st
 	for ix := range ports {
 		allowedPorts[ix] = strconv.Itoa(int(ports[ix].Port))
 	}
-	if !slicesEqual(allowedPorts, fw.Allowed[0].Ports) {
+	if !equalStringSets(allowedPorts, fw.Allowed[0].Ports) {
 		return true, true, nil
 	}
 	// The service controller already verified that the protocol matches on all ports, no need to check.
@@ -875,18 +938,47 @@ func (gce *GCECloud) firewallNeedsUpdate(name, serviceName, region, ipAddress st
 	return true, false, nil
 }
 
-func slicesEqual(x, y []string) bool {
-	if len(x) != len(y) {
-		return false
+func (gce *GCECloud) ensureHttpHealthCheckFirewall(serviceName, ipAddress, region string, hosts []*gceInstance, hcName string, hcPort int32, isNodesHealthCheck bool) error {
+	clusterID, err := gce.ClusterID.GetID()
+	if err != nil {
+		return fmt.Errorf("error getting cluster ID: %v", err)
 	}
-	sort.Strings(x)
-	sort.Strings(y)
-	for i := range x {
-		if x[i] != y[i] {
-			return false
+
+	// Prepare the firewall params for creating / checking.
+	desc := fmt.Sprintf(`{"kubernetes.io/cluster-id":"%s"}`, clusterID)
+	if !isNodesHealthCheck {
+		desc = makeFirewallDescription(serviceName, ipAddress)
+	}
+	sourceRanges := lbSrcRngsFlag.ipn
+	ports := []v1.ServicePort{{Protocol: "tcp", Port: hcPort}}
+
+	fwName := MakeHealthCheckFirewallName(clusterID, hcName, isNodesHealthCheck)
+	fw, err := gce.service.Firewalls.Get(gce.projectID, fwName).Do()
+	if err != nil {
+		if !isHTTPErrorCode(err, http.StatusNotFound) {
+			return fmt.Errorf("error getting firewall for health checks: %v", err)
 		}
+		glog.Infof("Creating firewall %v for health checks.", fwName)
+		if err := gce.createFirewall(fwName, region, desc, sourceRanges, ports, hosts); err != nil {
+			return err
+		}
+		glog.Infof("Created firewall %v for health checks.", fwName)
+		return nil
 	}
-	return true
+	// Validate firewall fields.
+	if fw.Description != desc ||
+		len(fw.Allowed) != 1 ||
+		fw.Allowed[0].IPProtocol != string(ports[0].Protocol) ||
+		!equalStringSets(fw.Allowed[0].Ports, []string{string(ports[0].Port)}) ||
+		!equalStringSets(fw.SourceRanges, sourceRanges.StringSlice()) {
+		glog.Warningf("Firewall %v exists but parameters have drifted - updating...", fwName)
+		if err := gce.updateFirewall(fwName, region, desc, sourceRanges, ports, hosts); err != nil {
+			glog.Warningf("Failed to reconcile firewall %v parameters.", fwName)
+			return err
+		}
+		glog.V(4).Infof("Corrected firewall %v parameters successful", fwName)
+	}
+	return nil
 }
 
 func (gce *GCECloud) createForwardingRule(name, serviceName, region, ipAddress string, ports []v1.ServicePort) error {
@@ -942,7 +1034,7 @@ func (gce *GCECloud) updateFirewall(name, region, desc string, sourceRanges nets
 	if err != nil {
 		return mc.Observe(err)
 	}
-	op, err := gce.service.Firewalls.Update(gce.projectID, makeFirewallName(name), firewall).Do()
+	op, err := gce.service.Firewalls.Update(gce.projectID, name, firewall).Do()
 	if err != nil && !isHTTPErrorCode(err, http.StatusConflict) {
 		return mc.Observe(err)
 	}
@@ -971,7 +1063,7 @@ func (gce *GCECloud) firewallObject(name, region, desc string, sourceRanges nets
 	}
 
 	firewall := &compute.Firewall{
-		Name:         makeFirewallName(name),
+		Name:         name,
 		Description:  desc,
 		Network:      gce.networkURL,
 		SourceRanges: sourceRanges.StringSlice(),
@@ -1141,17 +1233,16 @@ func (gce *GCECloud) ensureStaticIP(name, serviceName, region, existingIP string
 
 func (gce *GCECloud) deleteFirewall(name, region string) error {
 	mc := newFirewallMetricContext("delete", region)
-	fwName := makeFirewallName(name)
-	op, err := gce.service.Firewalls.Delete(gce.projectID, fwName).Do()
+	op, err := gce.service.Firewalls.Delete(gce.projectID, name).Do()
 
 	if err != nil && isHTTPErrorCode(err, http.StatusNotFound) {
 		glog.V(2).Infof("Firewall %s already deleted. Continuing to delete other resources.", name)
 	} else if err != nil {
-		glog.Warningf("Failed to delete firewall %s, got error %v", fwName, err)
+		glog.Warningf("Failed to delete firewall %s, got error %v", name, err)
 		return mc.Observe(err)
 	} else {
 		if err := gce.waitForGlobalOp(op, mc); err != nil {
-			glog.Warningf("Failed waiting for Firewall %s to be deleted.  Got error: %v", fwName, err)
+			glog.Warningf("Failed waiting for Firewall %s to be deleted.  Got error: %v", name, err)
 			return err
 		}
 	}
diff --git a/pkg/cloudprovider/providers/gce/gce_util.go b/pkg/cloudprovider/providers/gce/gce_util.go
index 8f59cd5c495..90b51759f22 100644
--- a/pkg/cloudprovider/providers/gce/gce_util.go
+++ b/pkg/cloudprovider/providers/gce/gce_util.go
@@ -19,10 +19,12 @@ package gce
 import (
 	"errors"
 	"fmt"
+	"net/http"
 	"regexp"
 	"strings"
 
 	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/apimachinery/pkg/util/sets"
 
 	"cloud.google.com/go/compute/metadata"
 	compute "google.golang.org/api/compute/v1"
@@ -105,6 +107,14 @@ func isHTTPErrorCode(err error, code int) bool {
 	return ok && apiErr.Code == code
 }
 
+func isInUsedByError(err error) bool {
+	apiErr, ok := err.(*googleapi.Error)
+	if !ok || apiErr.Code != http.StatusBadRequest {
+		return false
+	}
+	return strings.Contains(apiErr.Message, "being used by")
+}
+
 // splitProviderID splits a provider's id into core components.
 // A providerID is build out of '${ProviderName}://${project-id}/${zone}/${instance-name}'
 // See cloudprovider.GetInstanceProviderID.
@@ -115,3 +125,12 @@ func splitProviderID(providerID string) (project, zone, instance string, err err
 	}
 	return matches[1], matches[2], matches[3], nil
 }
+
+func equalStringSets(x, y []string) bool {
+	if len(x) != len(y) {
+		return false
+	}
+	xString := sets.NewString(x...)
+	yString := sets.NewString(y...)
+	return xString.Equal(yString)
+}
diff --git a/test/e2e/firewall.go b/test/e2e/firewall.go
index d6640e18084..0030dfa57a5 100644
--- a/test/e2e/firewall.go
+++ b/test/e2e/firewall.go
@@ -22,6 +22,7 @@ import (
 	"k8s.io/apimachinery/pkg/util/sets"
 	"k8s.io/kubernetes/pkg/api/v1"
 	"k8s.io/kubernetes/pkg/client/clientset_generated/clientset"
+	"k8s.io/kubernetes/pkg/cloudprovider"
 	gcecloud "k8s.io/kubernetes/pkg/cloudprovider/providers/gce"
 	"k8s.io/kubernetes/pkg/master/ports"
 	"k8s.io/kubernetes/test/e2e/framework"
@@ -45,13 +46,18 @@ var _ = framework.KubeDescribe("Firewall rule", func() {
 		gceCloud = cloudConfig.Provider.(*gcecloud.GCECloud)
 	})
 
-	// This test takes around 4 minutes to run
+	// This test takes around 6 minutes to run
 	It("[Slow] [Serial] should create valid firewall rules for LoadBalancer type service", func() {
 		ns := f.Namespace.Name
 		// This source ranges is just used to examine we have exact same things on LB firewall rules
 		firewallTestSourceRanges := []string{"0.0.0.0/1", "128.0.0.0/1"}
 		serviceName := "firewall-test-loadbalancer"
 
+		By("Getting cluster ID")
+		clusterID, err := framework.GetClusterID(cs)
+		Expect(err).NotTo(HaveOccurred())
+		framework.Logf("Got cluster ID: %v", clusterID)
+
 		jig := framework.NewServiceTestJig(cs, serviceName)
 		nodesNames := jig.GetNodesNames(framework.MaxNodesForEndpointsTests)
 		if len(nodesNames) <= 0 {
@@ -59,28 +65,52 @@ var _ = framework.KubeDescribe("Firewall rule", func() {
 		}
 		nodesSet := sets.NewString(nodesNames...)
 
-		// OnlyLocal service is needed to examine which exact nodes the requests are being forwarded to by the Load Balancer on GCE
-		By("Creating a LoadBalancer type service with ExternalTrafficPolicy=Local")
-		svc := jig.CreateOnlyLocalLoadBalancerService(ns, serviceName,
-			framework.LoadBalancerCreateTimeoutDefault, false, func(svc *v1.Service) {
-				svc.Spec.Ports = []v1.ServicePort{{Protocol: "TCP", Port: framework.FirewallTestHttpPort}}
-				svc.Spec.LoadBalancerSourceRanges = firewallTestSourceRanges
-			})
+		By("Creating a LoadBalancer type service with ExternalTrafficPolicy=Global")
+		svc := jig.CreateLoadBalancerService(ns, serviceName, framework.LoadBalancerCreateTimeoutDefault, func(svc *v1.Service) {
+			svc.Spec.Ports = []v1.ServicePort{{Protocol: "TCP", Port: framework.FirewallTestHttpPort}}
+			svc.Spec.LoadBalancerSourceRanges = firewallTestSourceRanges
+		})
 		defer func() {
 			jig.UpdateServiceOrFail(svc.Namespace, svc.Name, func(svc *v1.Service) {
 				svc.Spec.Type = v1.ServiceTypeNodePort
 				svc.Spec.LoadBalancerSourceRanges = nil
 			})
 			Expect(cs.Core().Services(svc.Namespace).Delete(svc.Name, nil)).NotTo(HaveOccurred())
+			By("Waiting for the local traffic health check firewall rule to be deleted")
+			localHCFwName := framework.MakeHealthCheckFirewallNameForLBService(clusterID, cloudprovider.GetLoadBalancerName(svc), false)
+			_, err := framework.WaitForFirewallRule(gceCloud, localHCFwName, false, framework.LoadBalancerCleanupTimeout)
+			Expect(err).NotTo(HaveOccurred())
 		}()
 		svcExternalIP := svc.Status.LoadBalancer.Ingress[0].IP
 
-		By("Checking if service's firewall rules are correct")
+		By("Checking if service's firewall rule is correct")
 		nodeTags := framework.GetInstanceTags(cloudConfig, nodesNames[0])
-		expFw := framework.ConstructFirewallForLBService(svc, nodeTags.Items)
-		fw, err := gceCloud.GetFirewall(expFw.Name)
+		lbFw := framework.ConstructFirewallForLBService(svc, nodeTags.Items)
+		fw, err := gceCloud.GetFirewall(lbFw.Name)
 		Expect(err).NotTo(HaveOccurred())
-		Expect(framework.VerifyFirewallRule(fw, expFw, cloudConfig.Network, false)).NotTo(HaveOccurred())
+		Expect(framework.VerifyFirewallRule(fw, lbFw, cloudConfig.Network, false)).NotTo(HaveOccurred())
+
+		By("Checking if service's nodes health check firewall rule is correct")
+		nodesHCFw := framework.ConstructHealthCheckFirewallForLBService(clusterID, svc, nodeTags.Items, true)
+		fw, err = gceCloud.GetFirewall(nodesHCFw.Name)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(framework.VerifyFirewallRule(fw, nodesHCFw, cloudConfig.Network, false)).NotTo(HaveOccurred())
+
+		// OnlyLocal service is needed to examine which exact nodes the requests are being forwarded to by the Load Balancer on GCE
+		By("Updating LoadBalancer service to ExternalTrafficPolicy=Local")
+		svc = jig.UpdateServiceOrFail(svc.Namespace, svc.Name, func(svc *v1.Service) {
+			svc.Spec.ExternalTrafficPolicy = v1.ServiceExternalTrafficPolicyTypeLocal
+		})
+
+		By("Waiting for the nodes health check firewall rule to be deleted")
+		_, err = framework.WaitForFirewallRule(gceCloud, nodesHCFw.Name, false, framework.LoadBalancerCleanupTimeout)
+		Expect(err).NotTo(HaveOccurred())
+
+		By("Waiting for the correct local traffic health check firewall rule to be created")
+		localHCFw := framework.ConstructHealthCheckFirewallForLBService(clusterID, svc, nodeTags.Items, false)
+		fw, err = framework.WaitForFirewallRule(gceCloud, localHCFw.Name, true, framework.LoadBalancerCreateTimeoutDefault)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(framework.VerifyFirewallRule(fw, localHCFw, cloudConfig.Network, false)).NotTo(HaveOccurred())
 
 		By(fmt.Sprintf("Creating netexec pods on at most %v nodes", framework.MaxNodesForEndpointsTests))
 		for i, nodeName := range nodesNames {
@@ -100,7 +130,7 @@ var _ = framework.KubeDescribe("Firewall rule", func() {
 		// by removing the tag on one vm and make sure it doesn't get any traffic. This is an imperfect
 		// simulation, we really want to check that traffic doesn't reach a vm outside the GKE cluster, but
 		// that's much harder to do in the current e2e framework.
-		By("Removing tags from one of the nodes")
+		By(fmt.Sprintf("Removing tags from one of the nodes: %v", nodesNames[0]))
 		nodesSet.Delete(nodesNames[0])
 		removedTags := framework.SetInstanceTags(cloudConfig, nodesNames[0], []string{})
 		defer func() {
diff --git a/test/e2e/framework/BUILD b/test/e2e/framework/BUILD
index 530287e91ab..8d96e4ea293 100644
--- a/test/e2e/framework/BUILD
+++ b/test/e2e/framework/BUILD
@@ -45,6 +45,7 @@ go_library(
         "//pkg/api/v1/helper:go_default_library",
         "//pkg/api/v1/node:go_default_library",
         "//pkg/api/v1/pod:go_default_library",
+        "//pkg/api/v1/service:go_default_library",
         "//pkg/apis/apps/v1beta1:go_default_library",
         "//pkg/apis/authorization/v1beta1:go_default_library",
         "//pkg/apis/batch:go_default_library",
diff --git a/test/e2e/framework/firewall_util.go b/test/e2e/framework/firewall_util.go
index fe6ede0a57d..2de5860777a 100644
--- a/test/e2e/framework/firewall_util.go
+++ b/test/e2e/framework/firewall_util.go
@@ -18,12 +18,16 @@ package framework
 
 import (
 	"fmt"
+	"net/http"
 	"strconv"
 	"strings"
 	"time"
 
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/sets"
+	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/kubernetes/pkg/api/v1"
+	apiservice "k8s.io/kubernetes/pkg/api/v1/service"
 	"k8s.io/kubernetes/pkg/client/clientset_generated/clientset"
 	"k8s.io/kubernetes/pkg/cloudprovider"
 	gcecloud "k8s.io/kubernetes/pkg/cloudprovider/providers/gce"
@@ -41,7 +45,7 @@ const (
 )
 
 // MakeFirewallNameForLBService return the expected firewall name for a LB service.
-// This should match the formatting of makeFirewallName() in pkg/cloudprovider/providers/gce/gce.go
+// This should match the formatting of makeFirewallName() in pkg/cloudprovider/providers/gce/gce_loadbalancer.go
 func MakeFirewallNameForLBService(name string) string {
 	return fmt.Sprintf("k8s-fw-%s", name)
 }
@@ -68,6 +72,32 @@ func ConstructFirewallForLBService(svc *v1.Service, nodesTags []string) *compute
 	return &fw
 }
 
+func MakeHealthCheckFirewallNameForLBService(clusterID, name string, isNodesHealthCheck bool) string {
+	return gcecloud.MakeHealthCheckFirewallName(clusterID, name, isNodesHealthCheck)
+}
+
+// ConstructHealthCheckFirewallForLBService returns the expected GCE firewall rule for a loadbalancer type service
+func ConstructHealthCheckFirewallForLBService(clusterID string, svc *v1.Service, nodesTags []string, isNodesHealthCheck bool) *compute.Firewall {
+	if svc.Spec.Type != v1.ServiceTypeLoadBalancer {
+		Failf("can not construct firewall rule for non-loadbalancer type service")
+	}
+	fw := compute.Firewall{}
+	fw.Name = MakeHealthCheckFirewallNameForLBService(clusterID, cloudprovider.GetLoadBalancerName(svc), isNodesHealthCheck)
+	fw.TargetTags = nodesTags
+	fw.SourceRanges = gcecloud.LoadBalancerSrcRanges()
+	healthCheckPort := gcecloud.GetNodesHealthCheckPort()
+	if !isNodesHealthCheck {
+		healthCheckPort = apiservice.GetServiceHealthCheckNodePort(svc)
+	}
+	fw.Allowed = []*compute.FirewallAllowed{
+		{
+			IPProtocol: "tcp",
+			Ports:      []string{fmt.Sprintf("%d", healthCheckPort)},
+		},
+	}
+	return &fw
+}
+
 // GetNodeTags gets tags from one of the Kubernetes nodes
 func GetNodeTags(c clientset.Interface, cloudConfig CloudConfig) *compute.Tags {
 	nodes := GetReadySchedulableNodesOrDie(c)
@@ -303,6 +333,9 @@ func SameStringArray(result, expected []string, include bool) error {
 // VerifyFirewallRule verifies whether the result firewall is consistent with the expected firewall.
 // When `portsSubset` is false, match given ports exactly. Otherwise, only check ports are included.
 func VerifyFirewallRule(res, exp *compute.Firewall, network string, portsSubset bool) error {
+	if res == nil || exp == nil {
+		return fmt.Errorf("res and exp must not be nil")
+	}
 	if res.Name != exp.Name {
 		return fmt.Errorf("incorrect name: %v, expected %v", res.Name, exp.Name)
 	}
@@ -325,3 +358,40 @@ func VerifyFirewallRule(res, exp *compute.Firewall, network string, portsSubset
 	}
 	return nil
 }
+
+func WaitForFirewallRule(gceCloud *gcecloud.GCECloud, fwName string, exist bool, timeout time.Duration) (*compute.Firewall, error) {
+	Logf("Waiting up to %v for firewall %v exist=%v", timeout, fwName, exist)
+	var fw *compute.Firewall
+	var err error
+
+	condition := func() (bool, error) {
+		fw, err = gceCloud.GetFirewall(fwName)
+		if err != nil && exist ||
+			err == nil && !exist ||
+			err != nil && !exist && !IsGoogleAPIHTTPErrorCode(err, http.StatusNotFound) {
+			return false, nil
+		}
+		return true, nil
+	}
+
+	if err := wait.PollImmediate(5*time.Second, timeout, condition); err != nil {
+		return nil, fmt.Errorf("error waiting for firewall %v exist=%v", fwName, exist)
+	}
+	return fw, nil
+}
+
+func GetClusterID(c clientset.Interface) (string, error) {
+	cm, err := c.Core().ConfigMaps(metav1.NamespaceSystem).Get(gcecloud.UIDConfigMapName, metav1.GetOptions{})
+	if err != nil || cm == nil {
+		return "", fmt.Errorf("error getting cluster ID: %v", err)
+	}
+	clusterID, clusterIDExists := cm.Data[gcecloud.UIDCluster]
+	providerID, providerIDExists := cm.Data[gcecloud.UIDProvider]
+	if !clusterIDExists {
+		return "", fmt.Errorf("cluster ID not set")
+	}
+	if providerIDExists {
+		return providerID, nil
+	}
+	return clusterID, nil
+}
diff --git a/test/e2e/framework/service_util.go b/test/e2e/framework/service_util.go
index e91cff78b07..238b7ae1e12 100644
--- a/test/e2e/framework/service_util.go
+++ b/test/e2e/framework/service_util.go
@@ -235,6 +235,25 @@ func (j *ServiceTestJig) CreateOnlyLocalLoadBalancerService(namespace, serviceNa
 	return svc
 }
 
+// CreateLoadBalancerService creates a loadbalancer service and waits
+// for it to acquire an ingress IP.
+func (j *ServiceTestJig) CreateLoadBalancerService(namespace, serviceName string, timeout time.Duration, tweak func(svc *v1.Service)) *v1.Service {
+	By("creating a service " + namespace + "/" + serviceName + " with type=LoadBalancer")
+	svc := j.CreateTCPServiceOrFail(namespace, func(svc *v1.Service) {
+		svc.Spec.Type = v1.ServiceTypeLoadBalancer
+		// We need to turn affinity off for our LB distribution tests
+		svc.Spec.SessionAffinity = v1.ServiceAffinityNone
+		if tweak != nil {
+			tweak(svc)
+		}
+	})
+
+	By("waiting for loadbalancer for service " + namespace + "/" + serviceName)
+	svc = j.WaitForLoadBalancerOrFail(namespace, serviceName, timeout)
+	j.SanityCheckService(svc, v1.ServiceTypeLoadBalancer)
+	return svc
+}
+
 func GetNodeAddresses(node *v1.Node, addressType v1.NodeAddressType) (ips []string) {
 	for j := range node.Status.Addresses {
 		nodeAddress := &node.Status.Addresses[j]
diff --git a/test/e2e/framework/util.go b/test/e2e/framework/util.go
index 32b1ebca95d..0f31c013ed9 100644
--- a/test/e2e/framework/util.go
+++ b/test/e2e/framework/util.go
@@ -5181,12 +5181,16 @@ func CleanupGCEResources(loadBalancerName string) (retErr error) {
 	if err := DeleteGCEStaticIP(loadBalancerName); err != nil {
 		Logf("%v", err)
 	}
+	var hcNames []string
 	hc, getErr := gceCloud.GetHttpHealthCheck(loadBalancerName)
 	if getErr != nil && !IsGoogleAPIHTTPErrorCode(getErr, http.StatusNotFound) {
 		retErr = fmt.Errorf("%v\n%v", retErr, getErr)
 		return
 	}
-	if err := gceCloud.DeleteTargetPool(loadBalancerName, hc); err != nil &&
+	if hc != nil {
+		hcNames = append(hcNames, hc.Name)
+	}
+	if err := gceCloud.DeleteTargetPool(loadBalancerName, hcNames...); err != nil &&
 		!IsGoogleAPIHTTPErrorCode(err, http.StatusNotFound) {
 		retErr = fmt.Errorf("%v\n%v", retErr, err)
 	}