node monitoring: correctly recognize NotReady Nodes as unscheduable (#345)

Author: dgrove-oss

internal/controller/appwrapper/node_health_monitor.go

@@ -56,7 +56,8 @@ var (
 	// noScheduleNodes is a mapping from Node names to ResourceLists of unschedulable resources.
 	// A resource may be unschedulable either because:
 	//  (a) the Node is cordoned (node.Spec.Unschedulable is true) or
-	//  (b) Autopilot has labeled the Node with a NoExecute or NoSchedule taint for the resource.
+	//  (b) the Node has been marked as NotReady by Kubernetes or
+	//  (c) Autopilot has labeled the Node with a NoExecute or NoSchedule taint for the resource.
 	noScheduleNodes = make(map[string]v1.ResourceList)
 	// noScheduleNodesMutex synchronizes access to noScheduleNodes
 	noScheduleNodesMutex sync.RWMutex
@@ -153,7 +154,7 @@ func (r *NodeHealthMonitor) updateNoExecuteNodes(ctx context.Context, node *v1.N
 // update noScheduleNodes entry for node
 func (r *NodeHealthMonitor) updateNoScheduleNodes(ctx context.Context, node *v1.Node) {
 	var noScheduleResources v1.ResourceList
-	if node.Spec.Unschedulable {
+	if r.nodeIsUnscheduable(node) {
 		noScheduleResources = node.Status.Capacity.DeepCopy()
 		delete(noScheduleResources, v1.ResourcePods)
 	} else {
@@ -196,6 +197,18 @@ func (r *NodeHealthMonitor) updateNoScheduleNodes(ctx context.Context, node *v1.
 	}
 }
 
+func (r *NodeHealthMonitor) nodeIsUnscheduable(node *v1.Node) bool {
+	if node.Spec.Unschedulable {
+		return true
+	}
+	for _, taint := range node.Spec.Taints {
+		if taint.Key == "node.kubernetes.io/unreachable" || taint.Key == "node.kubernetes.io/not-ready" {
+			return true
+		}
+	}
+	return false
+}
+
 // SetupWithManager sets up the controller with the Manager.
 func (r *NodeHealthMonitor) SetupWithManager(mgr ctrl.Manager) error {
 	return ctrl.NewControllerManagedBy(mgr).

internal/controller/appwrapper/node_health_monitor_test.go

@@ -45,7 +45,13 @@ var _ = Describe("NodeMonitor Controller", func() {
 		Expect(k8sClient.Create(ctx, node)).To(Succeed())
 		node = getNode(nodeName)
 		node.Status.Capacity = nodeGPUs
+		node.Status.Conditions = append(node.Status.Conditions, v1.NodeCondition{
+			Type:   v1.NodeReady,
+			Status: v1.ConditionTrue,
+		})
 		Expect(k8sClient.Status().Update(ctx, node)).To(Succeed())
+		node.Spec.Taints = []v1.Taint{}
+		Expect(k8sClient.Update(ctx, node)).To(Succeed())
 	}
 
 	deleteNode := func(nodeName string) {
@@ -117,6 +123,62 @@ var _ = Describe("NodeMonitor Controller", func() {
 		Expect(err).NotTo(HaveOccurred())
 		Expect(len(noExecuteNodes)).Should(Equal(0))
 
+		By("A Node tainted as unreachable is detected as unscheduable")
+		node = getNode(node1Name.Name)
+		node.Spec.Taints = append(node.Spec.Taints, v1.Taint{Key: "node.kubernetes.io/unreachable", Effect: v1.TaintEffectNoExecute})
+		Expect(k8sClient.Update(ctx, node)).Should(Succeed())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node2Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(HaveLen(1))
+		Expect(noScheduleNodes).Should(HaveKey(node1Name.Name))
+		Expect(noScheduleNodes[node1Name.Name]).Should(HaveKey(v1.ResourceName("nvidia.com/gpu")))
+
+		By("Repeated reconcile does not change map")
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node2Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(HaveLen(1))
+		Expect(noScheduleNodes).Should(HaveKey(node1Name.Name))
+		Expect(noScheduleNodes[node1Name.Name]).Should(HaveKey(v1.ResourceName("nvidia.com/gpu")))
+
+		By("Removing the taint updates unhealthyNodes")
+		node.Spec.Taints = []v1.Taint{}
+		Expect(k8sClient.Update(ctx, node)).Should(Succeed())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(BeEmpty())
+
+		By("A Node tainted as not-read is detected as unscheduable")
+		node = getNode(node1Name.Name)
+		node.Spec.Taints = append(node.Spec.Taints, v1.Taint{Key: "node.kubernetes.io/not-ready", Effect: v1.TaintEffectNoExecute})
+		Expect(k8sClient.Update(ctx, node)).Should(Succeed())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node2Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(HaveLen(1))
+		Expect(noScheduleNodes).Should(HaveKey(node1Name.Name))
+		Expect(noScheduleNodes[node1Name.Name]).Should(HaveKey(v1.ResourceName("nvidia.com/gpu")))
+
+		By("Repeated reconcile does not change map")
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node2Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(HaveLen(1))
+		Expect(noScheduleNodes).Should(HaveKey(node1Name.Name))
+		Expect(noScheduleNodes[node1Name.Name]).Should(HaveKey(v1.ResourceName("nvidia.com/gpu")))
+
+		By("Removing the taint updates unhealthyNodes")
+		node.Spec.Taints = []v1.Taint{}
+		Expect(k8sClient.Update(ctx, node)).Should(Succeed())
+		_, err = nodeMonitor.Reconcile(ctx, reconcile.Request{NamespacedName: node1Name})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(noScheduleNodes).Should(BeEmpty())
+
 		deleteNode(node1Name.Name)
 		deleteNode(node2Name.Name)
 	})