diff --git a/test/e2e_node/device_manager_test.go b/test/e2e_node/device_manager_test.go
index c290a02d7bd..4a1f9398ef8 100644
--- a/test/e2e_node/device_manager_test.go
+++ b/test/e2e_node/device_manager_test.go
@@ -401,7 +401,7 @@ var _ = SIGDescribe("Device Manager", framework.WithSerial(), feature.DeviceMana
 				Should(HaveAllocatableDevices())
 		})
 
-		ginkgo.It("should deploy pod consuming devices first but fail with admission error after kubelet restart in case device plugin hasn't re-registered", func(ctx context.Context) {
+		framework.It("should deploy pod consuming devices first but fail with admission error after kubelet restart in case device plugin hasn't re-registered", framework.WithFlaky(), func(ctx context.Context) {
 			var err error
 			podCMD := "while true; do sleep 1000; done;"
 
diff --git a/test/e2e_node/device_plugin_test.go b/test/e2e_node/device_plugin_test.go
index 41f348c6768..749dabc9e62 100644
--- a/test/e2e_node/device_plugin_test.go
+++ b/test/e2e_node/device_plugin_test.go
@@ -844,7 +844,7 @@ func testDevicePluginNodeReboot(f *framework.Framework, pluginSockDir string) {
 		// simulate node reboot scenario by removing pods using CRI before kubelet is started. In addition to that,
 		// intentionally a scenario is created where after node reboot, application pods requesting devices appear before the device plugin pod
 		// exposing those devices as resource has restarted. The expected behavior is that the application pod fails at admission time.
-		ginkgo.It("Keeps device plugin assignments across node reboots (no pod restart, no device plugin re-registration)", func(ctx context.Context) {
+		framework.It("Keeps device plugin assignments across node reboots (no pod restart, no device plugin re-registration)", framework.WithFlaky(), func(ctx context.Context) {
 			podRECMD := fmt.Sprintf("devs=$(ls /tmp/ | egrep '^Dev-[0-9]+$') && echo stub devices: $devs && sleep %s", sleepIntervalForever)
 			pod1 := e2epod.NewPodClient(f).CreateSync(ctx, makeBusyboxPod(SampleDeviceResourceName, podRECMD))
 			deviceIDRE := "stub devices: (Dev-[0-9]+)"
diff --git a/test/e2e_node/node_shutdown_linux_test.go b/test/e2e_node/node_shutdown_linux_test.go
index 3af1b2a9a7a..88e7f4d3fca 100644
--- a/test/e2e_node/node_shutdown_linux_test.go
+++ b/test/e2e_node/node_shutdown_linux_test.go
@@ -376,7 +376,7 @@ var _ = SIGDescribe("GracefulNodeShutdown", framework.WithSerial(), nodefeature.
 		})
 	})
 
-	ginkgo.Context("when gracefully shutting down with Pod priority", func() {
+	framework.Context("when gracefully shutting down with Pod priority", framework.WithFlaky(), func() {
 
 		const (
 			pollInterval                 = 1 * time.Second
diff --git a/test/e2e_node/podresources_test.go b/test/e2e_node/podresources_test.go
index ed3db638ac3..edeb4ca61fa 100644
--- a/test/e2e_node/podresources_test.go
+++ b/test/e2e_node/podresources_test.go
@@ -953,7 +953,7 @@ var _ = SIGDescribe("POD Resources", framework.WithSerial(), feature.PodResource
 		})
 	})
 
-	ginkgo.Context("without SRIOV devices in the system", func() {
+	framework.Context("without SRIOV devices in the system", framework.WithFlaky(), func() {
 		ginkgo.BeforeEach(func() {
 			requireLackOfSRIOVDevices()
 		})
@@ -1224,7 +1224,7 @@ var _ = SIGDescribe("POD Resources", framework.WithSerial(), feature.PodResource
 		})
 	})
 
-	ginkgo.Context("with the builtin rate limit values", func() {
+	framework.Context("with the builtin rate limit values", framework.WithFlaky(), func() {
 		ginkgo.It("should hit throttling when calling podresources List in a tight loop", func(ctx context.Context) {
 			// ensure APIs have been called at least once
 			endpoint, err := util.LocalEndpoint(defaultPodResourcesPath, podresources.Socket)