Merge pull request #72259 from bsalamat/fix_nominated_node

pkg/scheduler/core/generic_scheduler.go

@@ -362,7 +362,7 @@ func (g *genericScheduler) processPreemptionWithExtenders(
 // worth the complexity, especially because we generally expect to have a very
 // small number of nominated pods per node.
 func (g *genericScheduler) getLowerPriorityNominatedPods(pod *v1.Pod, nodeName string) []*v1.Pod {
-	pods := g.schedulingQueue.WaitingPodsForNode(nodeName)
+	pods := g.schedulingQueue.NominatedPodsForNode(nodeName)
 
 	if len(pods) == 0 {
 		return nil
@@ -509,7 +509,7 @@ func addNominatedPods(pod *v1.Pod, meta predicates.PredicateMetadata,
 		// This may happen only in tests.
 		return false, meta, nodeInfo
 	}
-	nominatedPods := queue.WaitingPodsForNode(nodeInfo.Node().Name)
+	nominatedPods := queue.NominatedPodsForNode(nodeInfo.Node().Name)
 	if nominatedPods == nil || len(nominatedPods) == 0 {
 		return false, meta, nodeInfo
 	}

pkg/scheduler/scheduler.go

@@ -318,11 +318,19 @@ func (sched *Scheduler) preempt(preemptor *v1.Pod, scheduleErr error) (string, e
 	var nodeName = ""
 	if node != nil {
 		nodeName = node.Name
+		// Update the scheduling queue with the nominated pod information. Without
+		// this, there would be a race condition between the next scheduling cycle
+		// and the time the scheduler receives a Pod Update for the nominated pod.
+		sched.config.SchedulingQueue.UpdateNominatedPodForNode(preemptor, nodeName)
+
+		// Make a call to update nominated node name of the pod on the API server.
 		err = sched.config.PodPreemptor.SetNominatedNodeName(preemptor, nodeName)
 		if err != nil {
 			klog.Errorf("Error in preemption process. Cannot update pod %v/%v annotations: %v", preemptor.Namespace, preemptor.Name, err)
+			sched.config.SchedulingQueue.DeleteNominatedPodIfExists(preemptor)
 			return "", err
 		}
+
 		for _, victim := range victims {
 			if err := sched.config.PodPreemptor.DeletePod(victim); err != nil {
 				klog.Errorf("Error preempting pod %v/%v: %v", victim.Namespace, victim.Name, err)

test/integration/scheduler/BUILD

@@ -26,6 +26,7 @@ go_test(
         "//cmd/kube-scheduler/app:go_default_library",
         "//cmd/kube-scheduler/app/config:go_default_library",
         "//pkg/api/legacyscheme:go_default_library",
+        "//pkg/api/v1/pod:go_default_library",
         "//pkg/controller/nodelifecycle:go_default_library",
         "//pkg/controller/volume/persistentvolume:go_default_library",
         "//pkg/controller/volume/persistentvolume/options:go_default_library",

test/integration/scheduler/util.go

@@ -584,6 +584,17 @@ func podScheduled(c clientset.Interface, podNamespace, podName string) wait.Cond
 
 // podUnschedulable returns a condition function that returns true if the given pod
 // gets unschedulable status.
+func podSchedulableCondition(c clientset.Interface, podNamespace, podName string) (*v1.PodCondition, error) {
+	pod, err := c.CoreV1().Pods(podNamespace).Get(podName, metav1.GetOptions{})
+	if err != nil {
+		return nil, err
+	}
+	_, cond := podutil.GetPodCondition(&pod.Status, v1.PodScheduled)
+	return cond, nil
+}
+
+// podUnschedulable returns a condition function that returns true if the given pod
+// gets unschedulable status.
 func podUnschedulable(c clientset.Interface, podNamespace, podName string) wait.ConditionFunc {
 	return func() (bool, error) {
 		pod, err := c.CoreV1().Pods(podNamespace).Get(podName, metav1.GetOptions{})
@@ -710,7 +721,7 @@ func cleanupPods(cs clientset.Interface, t *testing.T, pods []*v1.Pod) {
 		}
 	}
 	for _, p := range pods {
-		if err := wait.Poll(time.Second, wait.ForeverTestTimeout,
+		if err := wait.Poll(time.Millisecond, wait.ForeverTestTimeout,
 			podDeleted(cs, p.Namespace, p.Name)); err != nil {
 			t.Errorf("error while waiting for pod  %v/%v to get deleted: %v", p.Namespace, p.Name, err)
 		}