Create hostNetwork pods even if network plugin not ready

We do now admit pods (unlike the first attempt), but now we will stop
non-hostnetwork pods from starting if the network is not ready.

Issue #35409

pkg/kubelet/kubelet.go

@@ -1393,6 +1393,11 @@ func (kl *Kubelet) syncPod(o syncPodOptions) error {
 		return errOuter
 	}
 
+	// If the network plugin is not ready, only start the pod if it uses the host network
+	if rs := kl.runtimeState.networkErrors(); len(rs) != 0 && !podUsesHostNetwork(pod) {
+		return fmt.Errorf("network is not ready: %v", rs)
+	}
+
 	// Create Cgroups for the pod and apply resource parameters
 	// to them if cgroup-per-qos flag is enabled.
 	pcm := kl.containerManager.NewPodContainerManager()
@@ -1644,7 +1649,7 @@ func (kl *Kubelet) syncLoop(updates <-chan kubetypes.PodUpdate, handler SyncHand
 	defer housekeepingTicker.Stop()
 	plegCh := kl.pleg.Watch()
 	for {
-		if rs := kl.runtimeState.errors(); len(rs) != 0 {
+		if rs := kl.runtimeState.runtimeErrors(); len(rs) != 0 {
 			glog.Infof("skipping pod synchronization - %v", rs)
 			time.Sleep(5 * time.Second)
 			continue

pkg/kubelet/kubelet_node_status.go

@@ -577,7 +577,8 @@ func (kl *Kubelet) setNodeReadyCondition(node *api.Node) {
 	// ref: https://github.com/kubernetes/kubernetes/issues/16961
 	currentTime := unversioned.NewTime(kl.clock.Now())
 	var newNodeReadyCondition api.NodeCondition
-	if rs := kl.runtimeState.errors(); len(rs) == 0 {
+	rs := append(kl.runtimeState.runtimeErrors(), kl.runtimeState.networkErrors()...)
+	if len(rs) == 0 {
 		newNodeReadyCondition = api.NodeCondition{
 			Type:              api.NodeReady,
 			Status:            api.ConditionTrue,

pkg/kubelet/runonce_test.go

@@ -83,6 +83,7 @@ func TestRunOnce(t *testing.T) {
 		kubeClient:          &fake.Clientset{},
 		hostname:            testKubeletHostname,
 		nodeName:            testKubeletHostname,
+		runtimeState:        newRuntimeState(time.Second),
 	}
 	kb.containerManager = cm.NewStubContainerManager()
 

pkg/kubelet/runtime.go

@@ -68,16 +68,13 @@ func (s *runtimeState) setInitError(err error) {
 	s.initError = err
 }
 
-func (s *runtimeState) errors() []string {
+func (s *runtimeState) runtimeErrors() []string {
 	s.RLock()
 	defer s.RUnlock()
 	var ret []string
 	if s.initError != nil {
 		ret = append(ret, s.initError.Error())
 	}
-	if s.networkError != nil {
-		ret = append(ret, s.networkError.Error())
-	}
 	if !s.lastBaseRuntimeSync.Add(s.baseRuntimeSyncThreshold).After(time.Now()) {
 		ret = append(ret, "container runtime is down")
 	}
@@ -87,6 +84,16 @@ func (s *runtimeState) errors() []string {
 	return ret
 }
 
+func (s *runtimeState) networkErrors() []string {
+	s.RLock()
+	defer s.RUnlock()
+	var ret []string
+	if s.networkError != nil {
+		ret = append(ret, s.networkError.Error())
+	}
+	return ret
+}
+
 func newRuntimeState(
 	runtimeSyncThreshold time.Duration,
 ) *runtimeState {