diff --git a/test/e2e_node/container_lifecycle_test.go b/test/e2e_node/container_lifecycle_test.go
index 0e744c22c47..d52c2b304e1 100644
--- a/test/e2e_node/container_lifecycle_test.go
+++ b/test/e2e_node/container_lifecycle_test.go
@@ -1234,17 +1234,14 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 			podSandboxID := sandboxes[0].Id
 
 			ginkgo.By("Stopping the kubelet")
-			restartKubelet := stopKubelet()
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet was expected to be stopped but it is still running"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By("Stopping the pod sandbox to simulate the node reboot")
 			err = rs.StopPodSandbox(ctx, podSandboxID)
 			framework.ExpectNoError(err)
 
 			ginkgo.By("Restarting the kubelet")
-			restartKubelet()
+			restartKubelet(ctx)
 			gomega.Eventually(ctx, func() bool {
 				return kubeletHealthCheck(kubeletHealthCheckURL)
 			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("kubelet was expected to be healthy"))
@@ -1361,14 +1358,10 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 
 			ginkgo.It("should not restart any completed init container after the kubelet restart", func(ctx context.Context) {
 				ginkgo.By("stopping the kubelet")
-				startKubelet := stopKubelet()
-				// wait until the kubelet health check will fail
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet should be stopped"))
+				restartKubelet := mustStopKubelet(ctx, f)
 
 				ginkgo.By("restarting the kubelet")
-				startKubelet()
+				restartKubelet(ctx)
 				// wait until the kubelet health check will succeed
 				gomega.Eventually(ctx, func() bool {
 					return kubeletHealthCheck(kubeletHealthCheckURL)
@@ -1409,11 +1402,7 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 
 			ginkgo.It("should not restart any completed init container, even after the completed init container statuses have been removed and the kubelet restarted", func(ctx context.Context) {
 				ginkgo.By("stopping the kubelet")
-				startKubelet := stopKubelet()
-				// wait until the kubelet health check will fail
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet should be stopped"))
+				restartKubelet := mustStopKubelet(ctx, f)
 
 				ginkgo.By("removing the completed init container statuses from the container runtime")
 				rs, _, err := getCRIClient()
@@ -1437,7 +1426,7 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 				}
 
 				ginkgo.By("restarting the kubelet")
-				startKubelet()
+				restartKubelet(ctx)
 				// wait until the kubelet health check will succeed
 				gomega.Eventually(ctx, func() bool {
 					return kubeletHealthCheck(kubeletHealthCheckURL)
@@ -1544,18 +1533,10 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 
 			ginkgo.It("should not restart any completed init container after the kubelet restart", func(ctx context.Context) {
 				ginkgo.By("stopping the kubelet")
-				startKubelet := stopKubelet()
-				// wait until the kubelet health check will fail
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet should be stopped"))
+				restartKubelet := mustStopKubelet(ctx, f)
 
 				ginkgo.By("restarting the kubelet")
-				startKubelet()
-				// wait until the kubelet health check will succeed
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("kubelet should be restarted"))
+				restartKubelet(ctx)
 
 				ginkgo.By("ensuring that no completed init container is restarted")
 				gomega.Consistently(ctx, func() bool {
@@ -1588,11 +1569,7 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 
 			ginkgo.It("should not restart any completed init container, even after the completed init container statuses have been removed and the kubelet restarted", func(ctx context.Context) {
 				ginkgo.By("stopping the kubelet")
-				startKubelet := stopKubelet()
-				// wait until the kubelet health check will fail
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet should be stopped"))
+				restartKubelet := mustStopKubelet(ctx, f)
 
 				ginkgo.By("removing the completed init container statuses from the container runtime")
 				rs, _, err := getCRIClient()
@@ -1616,11 +1593,7 @@ var _ = SIGDescribe(framework.WithSerial(), "Containers Lifecycle", func() {
 				}
 
 				ginkgo.By("restarting the kubelet")
-				startKubelet()
-				// wait until the kubelet health check will succeed
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("kubelet should be restarted"))
+				restartKubelet(ctx)
 
 				ginkgo.By("ensuring that no completed init container is restarted")
 				gomega.Consistently(ctx, func() bool {
@@ -5517,17 +5490,14 @@ var _ = SIGDescribe(nodefeature.SidecarContainers, framework.WithSerial(), "Cont
 			podSandboxID := sandboxes[0].Id
 
 			ginkgo.By("Stopping the kubelet")
-			restartKubelet := stopKubelet()
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet would have been stopped but it is still running"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By("Stopping the pod sandbox to simulate the node reboot")
 			err = rs.StopPodSandbox(ctx, podSandboxID)
 			framework.ExpectNoError(err)
 
 			ginkgo.By("Restarting the kubelet")
-			restartKubelet()
+			restartKubelet(ctx)
 			gomega.Eventually(ctx, func() bool {
 				return kubeletHealthCheck(kubeletHealthCheckURL)
 			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("kubelet was expected to be healthy"))
@@ -5662,10 +5632,7 @@ var _ = SIGDescribe(nodefeature.SidecarContainers, framework.WithSerial(), "Cont
 				podSandboxID := sandboxes[0].Id
 
 				ginkgo.By("Stopping the kubelet")
-				restartKubelet := stopKubelet()
-				gomega.Eventually(ctx, func() bool {
-					return kubeletHealthCheck(kubeletHealthCheckURL)
-				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet would have been stopped but it is still running"))
+				restartKubelet := mustStopKubelet(ctx, f)
 
 				if nodeReboot {
 					ginkgo.By("Stopping the pod sandbox to simulate the node reboot")
@@ -5674,7 +5641,7 @@ var _ = SIGDescribe(nodefeature.SidecarContainers, framework.WithSerial(), "Cont
 				}
 
 				ginkgo.By("Restarting the kubelet")
-				restartKubelet()
+				restartKubelet(ctx)
 				gomega.Eventually(ctx, func() bool {
 					return kubeletHealthCheck(kubeletHealthCheckURL)
 				}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("kubelet was expected to be healthy"))
diff --git a/test/e2e_node/device_manager_test.go b/test/e2e_node/device_manager_test.go
index af608dc15d9..dc0f6ae7625 100644
--- a/test/e2e_node/device_manager_test.go
+++ b/test/e2e_node/device_manager_test.go
@@ -194,7 +194,7 @@ var _ = SIGDescribe("Device Manager", framework.WithSerial(), nodefeature.Device
 			framework.Logf("pod %s/%s running", testPod.Namespace, testPod.Name)
 
 			ginkgo.By("stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By("stopping all the local containers - using CRI")
 			rs, _, err := getCRIClient()
@@ -210,7 +210,7 @@ var _ = SIGDescribe("Device Manager", framework.WithSerial(), nodefeature.Device
 			}
 
 			ginkgo.By("restarting the kubelet")
-			startKubelet()
+			restartKubelet(ctx)
 
 			ginkgo.By("waiting for the kubelet to be ready again")
 			// Wait for the Kubelet to be ready.
diff --git a/test/e2e_node/device_plugin_test.go b/test/e2e_node/device_plugin_test.go
index 5520c78dfed..2db84ce2b80 100644
--- a/test/e2e_node/device_plugin_test.go
+++ b/test/e2e_node/device_plugin_test.go
@@ -211,7 +211,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 				e2epod.NewPodClient(f).DeleteSync(ctx, p.Name, metav1.DeleteOptions{}, 2*time.Minute)
 			}
 
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("Waiting for devices to become unavailable on the local node")
 			gomega.Eventually(ctx, func(ctx context.Context) bool {
@@ -365,7 +365,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			framework.Logf("testing pod: pre-restart  UID=%s namespace=%s name=%s ready=%v", pod1.UID, pod1.Namespace, pod1.Name, podutils.IsPodReady(pod1))
 
 			ginkgo.By("Restarting Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("Wait for node to be ready again")
 			e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, 5*time.Minute)
@@ -430,7 +430,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			gomega.Expect(devIDRestart1).To(gomega.Equal(devID1))
 
 			ginkgo.By("Restarting Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("Wait for node to be ready again")
 			e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, 5*time.Minute)
@@ -499,7 +499,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			gomega.Eventually(getNodeResourceValues, devicePluginGracefulTimeout, f.Timeouts.Poll).WithContext(ctx).WithArguments(SampleDeviceResourceName).Should(gomega.Equal(ResourceValue{Allocatable: 0, Capacity: int(expectedSampleDevsAmount)}))
 
 			ginkgo.By("Restarting Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("Wait for node to be ready again")
 			gomega.Expect(e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, 5*time.Minute)).To(gomega.Succeed())
@@ -587,7 +587,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			framework.ExpectNoError(err)
 
 			ginkgo.By("Restarting Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("Wait for node to be ready again")
 			e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, 5*time.Minute)
@@ -652,7 +652,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			framework.ExpectNoError(err)
 
 			ginkgo.By("stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			// wait until the kubelet health check will fail
 			gomega.Eventually(ctx, func() bool {
@@ -666,7 +666,7 @@ func testDevicePlugin(f *framework.Framework, pluginSockDir string) {
 			deletePodSyncByName(ctx, f, pod.Name)
 
 			framework.Logf("Starting the kubelet")
-			startKubelet()
+			restartKubelet(ctx)
 
 			// wait until the kubelet health check will succeed
 			gomega.Eventually(ctx, func() bool {
@@ -947,7 +947,7 @@ func testDevicePluginNodeReboot(f *framework.Framework, pluginSockDir string) {
 			framework.ExpectNoError(err)
 
 			ginkgo.By("stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By("stopping all the local containers - using CRI")
 			rs, _, err := getCRIClient()
@@ -963,7 +963,7 @@ func testDevicePluginNodeReboot(f *framework.Framework, pluginSockDir string) {
 			}
 
 			ginkgo.By("restarting the kubelet")
-			startKubelet()
+			restartKubelet(ctx)
 
 			ginkgo.By("Wait for node to be ready again")
 			e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, 5*time.Minute)
diff --git a/test/e2e_node/dra_test.go b/test/e2e_node/dra_test.go
index 72ce7013762..45644fd8b5c 100644
--- a/test/e2e_node/dra_test.go
+++ b/test/e2e_node/dra_test.go
@@ -107,7 +107,7 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			}
 
 			ginkgo.By("restarting Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			ginkgo.By("wait for Kubelet plugin re-registration")
 			gomega.Eventually(getNewCalls).WithTimeout(pluginRegistrationTimeout).Should(testdriver.BeRegistered)
@@ -129,16 +129,15 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 
 			// Stop Kubelet
 			ginkgo.By("stop kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 			pod := createTestObjects(ctx, f.ClientSet, getNodeName(ctx, f), f.Namespace.Name, "draclass", "external-claim", "drapod", true, []string{driverName})
 			// Pod must be in pending state
 			err := e2epod.WaitForPodCondition(ctx, f.ClientSet, f.Namespace.Name, pod.Name, "Pending", framework.PodStartShortTimeout, func(pod *v1.Pod) (bool, error) {
 				return pod.Status.Phase == v1.PodPending, nil
 			})
 			framework.ExpectNoError(err)
-			// Start Kubelet
 			ginkgo.By("restart kubelet")
-			startKubelet()
+			restartKubelet(ctx)
 			// Pod should succeed
 			err = e2epod.WaitForPodSuccessInNamespaceTimeout(ctx, f.ClientSet, pod.Name, f.Namespace.Name, framework.PodStartShortTimeout)
 			framework.ExpectNoError(err)
@@ -227,12 +226,12 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodePrepareResourcesFailed)
 
 			ginkgo.By("stop Kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			unsetNodePrepareResourcesFailureMode()
 
-			ginkgo.By("start Kubelet")
-			startKubelet()
+			ginkgo.By("restart Kubelet")
+			restartKubelet(ctx)
 
 			ginkgo.By("wait for NodePrepareResources call to succeed")
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodePrepareResourcesSucceeded)
@@ -254,12 +253,12 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodeUnprepareResourcesFailed)
 
 			ginkgo.By("stop Kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			unsetNodeUnprepareResourcesFailureMode()
 
-			ginkgo.By("start Kubelet")
-			startKubelet()
+			ginkgo.By("restart Kubelet")
+			restartKubelet(ctx)
 
 			ginkgo.By("wait for NodeUnprepareResources call to succeed")
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodeUnprepareResourcesSucceeded)
@@ -313,7 +312,7 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodeUnprepareResourcesFailed)
 
 			ginkgo.By("restart Kubelet")
-			stopKubelet()()
+			restartKubelet(ctx, true)
 
 			ginkgo.By("wait for NodeUnprepareResources call to fail")
 			gomega.Eventually(kubeletPlugin.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodeUnprepareResourcesFailed)
@@ -337,15 +336,15 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			framework.ExpectNoError(err)
 
 			ginkgo.By("stop Kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By("delete pod")
 			e2epod.DeletePodOrFail(ctx, f.ClientSet, f.Namespace.Name, pod.Name)
 
 			unblockNodePrepareResources()
 
-			ginkgo.By("start Kubelet")
-			startKubelet()
+			ginkgo.By("restart Kubelet")
+			restartKubelet(ctx)
 
 			calls := kubeletPlugin.CountCalls("/NodePrepareResources")
 			ginkgo.By("make sure NodePrepareResources is not called again")
@@ -447,7 +446,7 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			framework.ExpectNoError(err)
 
 			ginkgo.By("restart Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			unblockNodePrepareResources()
 
@@ -472,7 +471,7 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 			gomega.Eventually(kubeletPlugin2.GetGRPCCalls).WithTimeout(retryTestTimeout).Should(testdriver.NodePrepareResourcesSucceeded)
 
 			ginkgo.By("restart Kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			unblockNodeUnprepareResources()
 
@@ -500,10 +499,10 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 
 		f.It("must be removed on kubelet startup", f.WithDisruptive(), func(ctx context.Context) {
 			ginkgo.By("stop kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 			ginkgo.DeferCleanup(func() {
-				if startKubelet != nil {
-					startKubelet()
+				if restartKubelet != nil {
+					restartKubelet(ctx)
 				}
 			})
 
@@ -518,9 +517,9 @@ var _ = framework.SIGDescribe("node")("DRA", feature.DynamicResourceAllocation,
 
 			gomega.Consistently(ctx, listResources).WithTimeout(5*time.Second).Should(matchAll, "ResourceSlices without kubelet")
 
-			ginkgo.By("start kubelet")
-			startKubelet()
-			startKubelet = nil
+			ginkgo.By("restart kubelet")
+			restartKubelet(ctx)
+			restartKubelet = nil
 
 			ginkgo.By("wait for exactly the node's ResourceSlice to get deleted")
 			gomega.Eventually(ctx, listResources).Should(matchOtherNode, "ResourceSlices with kubelet")
diff --git a/test/e2e_node/hugepages_test.go b/test/e2e_node/hugepages_test.go
index c7090d4c93f..b20ca0683ad 100644
--- a/test/e2e_node/hugepages_test.go
+++ b/test/e2e_node/hugepages_test.go
@@ -222,7 +222,7 @@ var _ = SIGDescribe("HugePages", framework.WithSerial(), feature.HugePages, "[No
 		gomega.Expect(value.String()).To(gomega.Equal("9Mi"), "huge pages with size 3Mi should be supported")
 
 		ginkgo.By("restarting the node and verifying that huge pages with size 3Mi are not supported")
-		restartKubelet(true)
+		restartKubelet(ctx, true)
 
 		ginkgo.By("verifying that the hugepages-3Mi resource no longer is present")
 		gomega.Eventually(ctx, func() bool {
@@ -235,14 +235,14 @@ var _ = SIGDescribe("HugePages", framework.WithSerial(), feature.HugePages, "[No
 
 	ginkgo.It("should add resources for new huge page sizes on kubelet restart", func(ctx context.Context) {
 		ginkgo.By("Stopping kubelet")
-		startKubelet := stopKubelet()
+		restartKubelet := mustStopKubelet(ctx, f)
 		ginkgo.By(`Patching away support for hugepage resource "hugepages-2Mi"`)
 		patch := []byte(`[{"op": "remove", "path": "/status/capacity/hugepages-2Mi"}, {"op": "remove", "path": "/status/allocatable/hugepages-2Mi"}]`)
 		result := f.ClientSet.CoreV1().RESTClient().Patch(types.JSONPatchType).Resource("nodes").Name(framework.TestContext.NodeName).SubResource("status").Body(patch).Do(ctx)
 		framework.ExpectNoError(result.Error(), "while patching")
 
-		ginkgo.By("Starting kubelet again")
-		startKubelet()
+		ginkgo.By("Restarting kubelet again")
+		restartKubelet(ctx)
 
 		ginkgo.By("verifying that the hugepages-2Mi resource is present")
 		gomega.Eventually(ctx, func() bool {
@@ -352,7 +352,7 @@ var _ = SIGDescribe("HugePages", framework.WithSerial(), feature.HugePages, "[No
 			setHugepages(ctx)
 
 			ginkgo.By("restarting kubelet to pick up pre-allocated hugepages")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			waitForHugepages(ctx)
 
@@ -370,7 +370,7 @@ var _ = SIGDescribe("HugePages", framework.WithSerial(), feature.HugePages, "[No
 			releaseHugepages(ctx)
 
 			ginkgo.By("restarting kubelet to pick up pre-allocated hugepages")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
 			waitForHugepages(ctx)
 		})
diff --git a/test/e2e_node/image_gc_test.go b/test/e2e_node/image_gc_test.go
index 5a725e66875..6c27dc92b20 100644
--- a/test/e2e_node/image_gc_test.go
+++ b/test/e2e_node/image_gc_test.go
@@ -94,8 +94,7 @@ var _ = SIGDescribe("ImageGarbageCollect", framework.WithSerial(), framework.Wit
 
 			e2epod.NewPodClient(f).DeleteSync(ctx, pod.ObjectMeta.Name, metav1.DeleteOptions{}, e2epod.DefaultPodDeletionTimeout)
 
-			restartKubelet(true)
-			waitForKubeletToStart(ctx, f)
+			restartKubelet(ctx, true)
 
 			// Wait until the maxAge of the image after the kubelet is restarted to ensure it doesn't
 			// GC too early.
diff --git a/test/e2e_node/kubelet_config_dir_test.go b/test/e2e_node/kubelet_config_dir_test.go
index 1d541afb17a..9e0e59ef117 100644
--- a/test/e2e_node/kubelet_config_dir_test.go
+++ b/test/e2e_node/kubelet_config_dir_test.go
@@ -54,12 +54,7 @@ var _ = SIGDescribe("Kubelet Config", framework.WithSlow(), framework.WithSerial
 			framework.ExpectNoError(err)
 
 			ginkgo.By("Stopping the kubelet")
-			restartKubelet := stopKubelet()
-
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			configDir := framework.TestContext.KubeletConfigDropinDir
 
@@ -128,7 +123,7 @@ featureGates:
   DynamicResourceAllocation: true`)
 			framework.ExpectNoError(os.WriteFile(filepath.Join(configDir, "20-kubelet.conf"), contents, 0755))
 			ginkgo.By("Restarting the kubelet")
-			restartKubelet()
+			restartKubelet(ctx)
 			// wait until the kubelet health check will succeed
 			gomega.Eventually(ctx, func() bool {
 				return kubeletHealthCheck(kubeletHealthCheckURL)
diff --git a/test/e2e_node/node_container_manager_test.go b/test/e2e_node/node_container_manager_test.go
index 6761e195a3b..5f61e4a8676 100644
--- a/test/e2e_node/node_container_manager_test.go
+++ b/test/e2e_node/node_container_manager_test.go
@@ -101,12 +101,9 @@ var _ = SIGDescribe("Node Container Manager", framework.WithSerial(), func() {
 					framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(oldCfg))
 
 					ginkgo.By("Restarting the kubelet")
-					restartKubelet(true)
+					restartKubelet(ctx, true)
 
-					// wait until the kubelet health check will succeed
-					gomega.Eventually(ctx, func(ctx context.Context) bool {
-						return kubeletHealthCheck(kubeletHealthCheckURL)
-					}).WithTimeout(2 * time.Minute).WithPolling(5 * time.Second).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+					waitForKubeletToStart(ctx, f)
 					ginkgo.By("Started the kubelet")
 				}
 			})
@@ -121,12 +118,9 @@ var _ = SIGDescribe("Node Container Manager", framework.WithSerial(), func() {
 			framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(newCfg))
 
 			ginkgo.By("Restarting the kubelet")
-			restartKubelet(true)
+			restartKubelet(ctx, true)
 
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func(ctx context.Context) bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}).WithTimeout(2 * time.Minute).WithPolling(5 * time.Second).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+			waitForKubeletToStart(ctx, f)
 			ginkgo.By("Started the kubelet")
 
 			gomega.Consistently(ctx, func(ctx context.Context) bool {
@@ -243,7 +237,7 @@ func runTest(ctx context.Context, f *framework.Framework) error {
 		if oldCfg != nil {
 			// Update the Kubelet configuration.
 			ginkgo.By("Stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			// wait until the kubelet health check will fail
 			gomega.Eventually(ctx, func() bool {
@@ -252,8 +246,8 @@ func runTest(ctx context.Context, f *framework.Framework) error {
 
 			framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(oldCfg))
 
-			ginkgo.By("Starting the kubelet")
-			startKubelet()
+			ginkgo.By("Restarting the kubelet")
+			restartKubelet(ctx)
 
 			// wait until the kubelet health check will succeed
 			gomega.Eventually(ctx, func(ctx context.Context) bool {
@@ -271,12 +265,7 @@ func runTest(ctx context.Context, f *framework.Framework) error {
 	// Set the new kubelet configuration.
 	// Update the Kubelet configuration.
 	ginkgo.By("Stopping the kubelet")
-	startKubelet := stopKubelet()
-
-	// wait until the kubelet health check will fail
-	gomega.Eventually(ctx, func() bool {
-		return kubeletHealthCheck(kubeletHealthCheckURL)
-	}, time.Minute, time.Second).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+	restartKubelet := mustStopKubelet(ctx, f)
 
 	expectedNAPodCgroup := cm.NewCgroupName(cm.RootCgroupName, nodeAllocatableCgroup)
 
@@ -293,7 +282,7 @@ func runTest(ctx context.Context, f *framework.Framework) error {
 	framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(newCfg))
 
 	ginkgo.By("Starting the kubelet")
-	startKubelet()
+	restartKubelet(ctx)
 
 	// wait until the kubelet health check will succeed
 	gomega.Eventually(ctx, func() bool {
diff --git a/test/e2e_node/node_perf_test.go b/test/e2e_node/node_perf_test.go
index 8f74c7ebc57..08a02b4ff3d 100644
--- a/test/e2e_node/node_perf_test.go
+++ b/test/e2e_node/node_perf_test.go
@@ -52,22 +52,12 @@ func setKubeletConfig(ctx context.Context, f *framework.Framework, cfg *kubeletc
 	if cfg != nil {
 		// Update the Kubelet configuration.
 		ginkgo.By("Stopping the kubelet")
-		startKubelet := stopKubelet()
-
-		// wait until the kubelet health check will fail
-		gomega.Eventually(ctx, func() bool {
-			return kubeletHealthCheck(kubeletHealthCheckURL)
-		}, time.Minute, time.Second).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+		restartKubelet := mustStopKubelet(ctx, f)
 
 		framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(cfg))
 
-		ginkgo.By("Starting the kubelet")
-		startKubelet()
-
-		// wait until the kubelet health check will succeed
-		gomega.Eventually(ctx, func() bool {
-			return kubeletHealthCheck(kubeletHealthCheckURL)
-		}, 2*time.Minute, 5*time.Second).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+		ginkgo.By("Restarting the kubelet")
+		restartKubelet(ctx)
 	}
 
 	// Wait for the Kubelet to be ready.
diff --git a/test/e2e_node/os_label_rename_test.go b/test/e2e_node/os_label_rename_test.go
index dbc456895f9..cb868d45d8e 100644
--- a/test/e2e_node/os_label_rename_test.go
+++ b/test/e2e_node/os_label_rename_test.go
@@ -49,7 +49,7 @@ var _ = SIGDescribe("OSArchLabelReconciliation", framework.WithSerial(), framewo
 
 			ginkgo.By("killing and restarting kubelet")
 			// Let's kill the kubelet
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 			// Update labels
 			newNode := node.DeepCopy()
 			newNode.Labels[v1.LabelOSStable] = "dummyOS"
@@ -57,7 +57,7 @@ var _ = SIGDescribe("OSArchLabelReconciliation", framework.WithSerial(), framewo
 			_, _, err := nodeutil.PatchNodeStatus(f.ClientSet.CoreV1(), types.NodeName(node.Name), node, newNode)
 			framework.ExpectNoError(err)
 			// Restart kubelet
-			startKubelet()
+			restartKubelet(ctx)
 			framework.ExpectNoError(e2enode.WaitForAllNodesSchedulable(ctx, f.ClientSet, framework.RestartNodeReadyAgainTimeout))
 			// If this happens right, node should have all the labels reset properly
 			err = waitForNodeLabels(ctx, f.ClientSet.CoreV1(), node.Name, 5*time.Minute)
diff --git a/test/e2e_node/podresources_test.go b/test/e2e_node/podresources_test.go
index b29eb5fba65..86e1573c2b5 100644
--- a/test/e2e_node/podresources_test.go
+++ b/test/e2e_node/podresources_test.go
@@ -1146,7 +1146,7 @@ var _ = SIGDescribe("POD Resources", framework.WithSerial(), feature.PodResource
 					expectPodResources(ctx, 1, cli, []podDesc{desc})
 
 					ginkgo.By("Restarting Kubelet")
-					restartKubelet(true)
+					restartKubelet(ctx, true)
 
 					// we need to wait for the node to be reported ready before we can safely query
 					// the podresources endpoint again. Otherwise we will have false negatives.
diff --git a/test/e2e_node/restart_test.go b/test/e2e_node/restart_test.go
index 2ea929e3747..ca3224f42d4 100644
--- a/test/e2e_node/restart_test.go
+++ b/test/e2e_node/restart_test.go
@@ -200,8 +200,8 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 
 			ginkgo.By("killing and restarting kubelet")
 			// We want to kill the kubelet rather than a graceful restart
-			startKubelet := stopKubelet()
-			startKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
+			restartKubelet(ctx)
 
 			// If this test works correctly, each of these pods will exit
 			// with no issue. But if accounting breaks, pods scheduled after
@@ -309,19 +309,10 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 			// As soon as the pod enters succeeded phase (detected by the watch above); kill the kubelet.
 			// This is a bit racy, but the goal is to stop the kubelet before the kubelet is able to delete the pod from the API-sever in order to repro https://issues.k8s.io/116925
 			ginkgo.By("Stopping the kubelet")
-			startKubelet := stopKubelet()
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
-			ginkgo.By("Starting the kubelet")
-			startKubelet()
-
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+			ginkgo.By("Restarting the kubelet")
+			restartKubelet(ctx)
 
 			// Wait for the Kubelet to be ready.
 			gomega.Eventually(ctx, func(ctx context.Context) bool {
@@ -361,12 +352,7 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 				},
 			})
 			ginkgo.By("Stopping the kubelet")
-			startKubelet := stopKubelet()
-
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			// Create the pod bound to the node. It will remain in the Pending
 			// phase as Kubelet is down.
@@ -379,19 +365,7 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 
 			// Restart Kubelet so that it proceeds with deletion
 			ginkgo.By("Starting the kubelet")
-			startKubelet()
-
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
-
-			// Wait for the Kubelet to be ready.
-			gomega.Eventually(ctx, func(ctx context.Context) bool {
-				nodes, err := e2enode.TotalReady(ctx, f.ClientSet)
-				framework.ExpectNoError(err)
-				return nodes == 1
-			}, time.Minute, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in ready state"))
+			restartKubelet(ctx)
 
 			ginkgo.By(fmt.Sprintf("After the kubelet is restarted, verify the pod (%v/%v) is deleted by kubelet", pod.Namespace, pod.Name))
 			gomega.Eventually(ctx, func(ctx context.Context) error {
@@ -439,12 +413,7 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 			framework.ExpectNoError(err, "Failed to await for the pod to be running: (%v/%v)", f.Namespace.Name, pod.Name)
 
 			ginkgo.By("Stopping the kubelet")
-			startKubelet := stopKubelet()
-
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			ginkgo.By(fmt.Sprintf("Deleting the pod (%v/%v) to set a deletion timestamp", pod.Namespace, pod.Name))
 			err = e2epod.NewPodClient(f).Delete(ctx, pod.Name, metav1.DeleteOptions{GracePeriodSeconds: &gracePeriod})
@@ -454,13 +423,8 @@ var _ = SIGDescribe("Restart", framework.WithSerial(), framework.WithSlow(), fra
 			e2enode.RemoveLabelOffNode(f.ClientSet, nodeName, nodeLabelKey)
 
 			// Restart Kubelet so that it proceeds with deletion
-			ginkgo.By("Starting the kubelet")
-			startKubelet()
-
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+			ginkgo.By("Restarting the kubelet")
+			restartKubelet(ctx)
 
 			// Wait for the Kubelet to be ready.
 			gomega.Eventually(ctx, func(ctx context.Context) bool {
diff --git a/test/e2e_node/unknown_pods_test.go b/test/e2e_node/unknown_pods_test.go
index 378b1e8ba76..436b696cf6e 100644
--- a/test/e2e_node/unknown_pods_test.go
+++ b/test/e2e_node/unknown_pods_test.go
@@ -69,29 +69,19 @@ var _ = SIGDescribe("Unknown Pods", framework.WithSerial(), framework.WithDisrup
 
 		ginkgo.It("the static pod should be terminated and cleaned up due to becoming a unknown pod due to being force deleted while kubelet is not running", func(ctx context.Context) {
 			framework.Logf("Stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			pod, err := f.ClientSet.CoreV1().Pods(ns).Get(ctx, mirrorPodName, metav1.GetOptions{})
 			framework.ExpectNoError(err)
 
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
-
 			framework.Logf("Delete the static pod manifest while the kubelet is not running")
 			file := staticPodPath(podPath, staticPodName, ns)
 			framework.Logf("deleting static pod manifest %q", file)
 			err = os.Remove(file)
 			framework.ExpectNoError(err)
 
-			framework.Logf("Starting the kubelet")
-			startKubelet()
-
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+			framework.Logf("Restarting the kubelet")
+			restartKubelet(ctx)
 
 			framework.Logf("wait for the mirror pod %v to disappear", mirrorPodName)
 			gomega.Eventually(ctx, func(ctx context.Context) error {
@@ -140,27 +130,17 @@ var _ = SIGDescribe("Unknown Pods", framework.WithSerial(), framework.WithDisrup
 
 		ginkgo.It("the api pod should be terminated and cleaned up due to becoming a unknown pod due to being force deleted while kubelet is not running", func(ctx context.Context) {
 			framework.Logf("Stopping the kubelet")
-			startKubelet := stopKubelet()
+			restartKubelet := mustStopKubelet(ctx, f)
 
 			pod, err := f.ClientSet.CoreV1().Pods(ns).Get(ctx, podName, metav1.GetOptions{})
 			framework.ExpectNoError(err)
 
-			// wait until the kubelet health check will fail
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
-
 			framework.Logf("Delete the pod while the kubelet is not running")
 			// Delete pod sync by name will force delete the pod, removing it from kubelet's config
 			deletePodSyncByName(ctx, f, podName)
 
-			framework.Logf("Starting the kubelet")
-			startKubelet()
-
-			// wait until the kubelet health check will succeed
-			gomega.Eventually(ctx, func() bool {
-				return kubeletHealthCheck(kubeletHealthCheckURL)
-			}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeTrueBecause("expected kubelet to be in healthy state"))
+			framework.Logf("Restarting the kubelet")
+			restartKubelet(ctx)
 
 			framework.Logf("wait for the pod %v to disappear", podName)
 			gomega.Eventually(ctx, func(ctx context.Context) error {
diff --git a/test/e2e_node/util.go b/test/e2e_node/util.go
index 6fc0e0a2e35..7919f13220b 100644
--- a/test/e2e_node/util.go
+++ b/test/e2e_node/util.go
@@ -216,12 +216,7 @@ func tempSetCurrentKubeletConfig(f *framework.Framework, updateFunction func(ctx
 func updateKubeletConfig(ctx context.Context, f *framework.Framework, kubeletConfig *kubeletconfig.KubeletConfiguration, deleteStateFiles bool) {
 	// Update the Kubelet configuration.
 	ginkgo.By("Stopping the kubelet")
-	startKubelet := stopKubelet()
-
-	// wait until the kubelet health check will fail
-	gomega.Eventually(ctx, func() bool {
-		return kubeletHealthCheck(kubeletHealthCheckURL)
-	}, time.Minute, time.Second).Should(gomega.BeFalseBecause("expected kubelet health check to be failed"))
+	restartKubelet := mustStopKubelet(ctx, f)
 
 	// Delete CPU and memory manager state files to be sure it will not prevent the kubelet restart
 	if deleteStateFiles {
@@ -231,9 +226,8 @@ func updateKubeletConfig(ctx context.Context, f *framework.Framework, kubeletCon
 
 	framework.ExpectNoError(e2enodekubelet.WriteKubeletConfigFile(kubeletConfig))
 
-	ginkgo.By("Starting the kubelet")
-	startKubelet()
-	waitForKubeletToStart(ctx, f)
+	ginkgo.By("Restarting the kubelet")
+	restartKubelet(ctx)
 }
 
 func waitForKubeletToStart(ctx context.Context, f *framework.Framework) {
@@ -433,31 +427,37 @@ func startContainerRuntime() error {
 // Warning: the "current" kubelet is poorly defined. The "current" kubelet is assumed to be the most
 // recent kubelet service unit, IOW there is not a unique ID we use to bind explicitly a kubelet
 // instance to a test run.
-func restartKubelet(running bool) {
+func restartKubelet(ctx context.Context, running bool) {
 	kubeletServiceName := findKubeletServiceName(running)
 	// reset the kubelet service start-limit-hit
-	stdout, err := exec.Command("sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()
+	stdout, err := exec.CommandContext(ctx, "sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()
 	framework.ExpectNoError(err, "Failed to reset kubelet start-limit-hit with systemctl: %v, %s", err, string(stdout))
 
-	stdout, err = exec.Command("sudo", "systemctl", "restart", kubeletServiceName).CombinedOutput()
+	stdout, err = exec.CommandContext(ctx, "sudo", "systemctl", "restart", kubeletServiceName).CombinedOutput()
 	framework.ExpectNoError(err, "Failed to restart kubelet with systemctl: %v, %s", err, string(stdout))
 }
 
-// stopKubelet will kill the running kubelet, and returns a func that will restart the process again
-func stopKubelet() func() {
+// mustStopKubelet will kill the running kubelet, and returns a func that will restart the process again
+func mustStopKubelet(ctx context.Context, f *framework.Framework) func(ctx context.Context) {
 	kubeletServiceName := findKubeletServiceName(true)
 
 	// reset the kubelet service start-limit-hit
-	stdout, err := exec.Command("sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()
+	stdout, err := exec.CommandContext(ctx, "sudo", "systemctl", "reset-failed", kubeletServiceName).CombinedOutput()
 	framework.ExpectNoError(err, "Failed to reset kubelet start-limit-hit with systemctl: %v, %s", err, string(stdout))
 
-	stdout, err = exec.Command("sudo", "systemctl", "kill", kubeletServiceName).CombinedOutput()
+	stdout, err = exec.CommandContext(ctx, "sudo", "systemctl", "kill", kubeletServiceName).CombinedOutput()
 	framework.ExpectNoError(err, "Failed to stop kubelet with systemctl: %v, %s", err, string(stdout))
 
-	return func() {
+	// wait until the kubelet health check fail
+	gomega.Eventually(ctx, func() bool {
+		return kubeletHealthCheck(kubeletHealthCheckURL)
+	}, f.Timeouts.PodStart, f.Timeouts.Poll).Should(gomega.BeFalseBecause("kubelet was expected to be stopped but it is still running"))
+
+	return func(ctx context.Context) {
 		// we should restart service, otherwise the transient service start will fail
-		stdout, err := exec.Command("sudo", "systemctl", "restart", kubeletServiceName).CombinedOutput()
+		stdout, err := exec.CommandContext(ctx, "sudo", "systemctl", "restart", kubeletServiceName).CombinedOutput()
 		framework.ExpectNoError(err, "Failed to restart kubelet with systemctl: %v, %v", err, stdout)
+		waitForKubeletToStart(ctx, f)
 	}
 }