cherry-pick: don't include terminated nodes in budget (#1735)

pkg/controllers/disruption/helpers.go

@@ -212,6 +212,13 @@ func BuildDisruptionBudgets(ctx context.Context, cluster *state.Cluster, clk clo
 			continue
 		}
 
+		// Additionally, don't consider nodeclaims that have the terminating condition. A nodeclaim should have
+		// the Terminating condition only when the node is drained and cloudprovider.Delete() was successful
+		// on the underlying cloud provider machine.
+		if node.NodeClaim.StatusConditions().Get(v1.ConditionTypeInstanceTerminating).IsTrue() {
+			continue
+		}
+
 		nodePool := node.Labels()[v1.NodePoolLabelKey]
 		numNodes[nodePool]++
 

pkg/controllers/disruption/suite_test.go

@@ -667,6 +667,38 @@ var _ = Describe("BuildDisruptionBudgetMapping", func() {
 			Expect(budgets[nodePool.Name][reason]).To(Equal(10))
 		}
 	})
+	It("should not consider nodes that have the terminating status condition as part of disruption count", func() {
+		nodePool.Spec.Disruption.Budgets = []v1.Budget{{Nodes: "100%"}}
+		ExpectApplied(ctx, env.Client, nodePool)
+		nodeClaim, node := test.NodeClaimAndNode(v1.NodeClaim{
+			ObjectMeta: metav1.ObjectMeta{
+				Finalizers: []string{"karpenter.sh/test-finalizer"},
+				Labels: map[string]string{
+					v1.NodePoolLabelKey:            nodePool.Name,
+					corev1.LabelInstanceTypeStable: mostExpensiveInstance.Name,
+					v1.CapacityTypeLabelKey:        mostExpensiveOffering.Requirements.Get(v1.CapacityTypeLabelKey).Any(),
+					corev1.LabelTopologyZone:       mostExpensiveOffering.Requirements.Get(corev1.LabelTopologyZone).Any(),
+				},
+			},
+			Status: v1.NodeClaimStatus{
+				Allocatable: map[corev1.ResourceName]resource.Quantity{
+					corev1.ResourceCPU:  resource.MustParse("32"),
+					corev1.ResourcePods: resource.MustParse("100"),
+				},
+			},
+		})
+		nodeClaim.StatusConditions().SetTrue(v1.ConditionTypeInstanceTerminating)
+		ExpectApplied(ctx, env.Client, nodeClaim, node)
+		ExpectReconcileSucceeded(ctx, nodeStateController, client.ObjectKeyFromObject(node))
+		ExpectReconcileSucceeded(ctx, nodeClaimStateController, client.ObjectKeyFromObject(nodeClaim))
+
+		budgets, err := disruption.BuildDisruptionBudgets(ctx, cluster, fakeClock, env.Client, recorder)
+		Expect(err).To(Succeed())
+		// This should not bring in the terminating node.
+		for _, reason := range v1.WellKnownDisruptionReasons {
+			Expect(budgets[nodePool.Name][reason]).To(Equal(10))
+		}
+	})
 	It("should not return a negative disruption value", func() {
 		nodePool.Spec.Disruption.Budgets = []v1.Budget{{Nodes: "10%"}}
 		ExpectApplied(ctx, env.Client, nodePool)

pkg/controllers/state/statenode.go

@@ -408,10 +408,10 @@ func (in *StateNode) PodLimits() corev1.ResourceList {
 func (in *StateNode) MarkedForDeletion() bool {
 	// The Node is marked for deletion if:
 	//  1. The Node has MarkedForDeletion set
-	//  2. The Node has a NodeClaim counterpart and is actively deleting
+	//  2. The Node has a NodeClaim counterpart and is actively deleting (or the nodeclaim is marked as terminating)
 	//  3. The Node has no NodeClaim counterpart and is actively deleting
 	return in.markedForDeletion ||
-		(in.NodeClaim != nil && !in.NodeClaim.DeletionTimestamp.IsZero()) ||
+		(in.NodeClaim != nil && (!in.NodeClaim.DeletionTimestamp.IsZero() || in.NodeClaim.StatusConditions().Get(v1.ConditionTypeInstanceTerminating).IsTrue())) ||
 		(in.Node != nil && in.NodeClaim == nil && !in.Node.DeletionTimestamp.IsZero())
 }