Merge pull request #48970 from caseydavenport/fix-kubelet-restart

pkg/kubelet/kuberuntime/kuberuntime_manager.go

@@ -396,6 +396,12 @@ func (m *kubeGenericRuntimeManager) podSandboxChanged(pod *v1.Pod, podStatus *ku
 		return true, sandboxStatus.Metadata.Attempt + 1, ""
 	}
 
+	// Needs to create a new sandbox when the sandbox does not have an IP address.
+	if !kubecontainer.IsHostNetworkPod(pod) && sandboxStatus.Network.Ip == "" {
+		glog.V(2).Infof("Sandbox for pod %q has no IP address.  Need to start a new one", format.Pod(pod))
+		return true, sandboxStatus.Metadata.Attempt + 1, sandboxStatus.Id
+	}
+
 	return false, sandboxStatus.Metadata.Attempt, sandboxStatus.Id
 }
 

pkg/kubelet/kuberuntime/kuberuntime_manager_test.go

@@ -750,6 +750,7 @@ func makeBasePodAndStatus() (*v1.Pod, *kubecontainer.PodStatus) {
 				Id:       "sandboxID",
 				State:    runtimeapi.PodSandboxState_SANDBOX_READY,
 				Metadata: &runtimeapi.PodSandboxMetadata{Name: pod.Name, Namespace: pod.Namespace, Uid: "sandboxuid", Attempt: uint32(0)},
+				Network:  &runtimeapi.PodSandboxNetworkStatus{Ip: "10.0.0.1"},
 			},
 		},
 		ContainerStatuses: []*kubecontainer.ContainerStatus{
@@ -883,7 +884,19 @@ func TestComputePodActions(t *testing.T) {
 				ContainersToKill:  getKillMap(basePod, baseStatus, []int{}),
 			},
 		},
-
+		"Kill pod and recreate all containers if the PodSandbox does not have an IP": {
+			mutateStatusFn: func(status *kubecontainer.PodStatus) {
+				status.SandboxStatuses[0].Network.Ip = ""
+			},
+			actions: podActions{
+				KillPod:           true,
+				CreateSandbox:     true,
+				SandboxID:         baseStatus.SandboxStatuses[0].Id,
+				Attempt:           uint32(1),
+				ContainersToStart: []int{0, 1, 2},
+				ContainersToKill:  getKillMap(basePod, baseStatus, []int{}),
+			},
+		},
 		"Kill and recreate the container if the container's spec changed": {
 			mutatePodFn: func(pod *v1.Pod) {
 				pod.Spec.RestartPolicy = v1.RestartPolicyAlways

pkg/kubelet/pleg/generic.go

@@ -177,7 +177,7 @@ func (g *GenericPLEG) updateRelisTime(timestamp time.Time) {
 }
 
 // relist queries the container runtime for list of pods/containers, compare
-// with the internal pods/containers, and generats events accordingly.
+// with the internal pods/containers, and generates events accordingly.
 func (g *GenericPLEG) relist() {
 	glog.V(5).Infof("GenericPLEG: Relisting")