Handle cluster status reconciliation with autoscaling configured

pkg/api/v1alpha1/condition_consts.go

@@ -55,6 +55,9 @@ const (
 
 	// ExternalEtcdNotAvailable reports the Cluster status is waiting for Etcd to be available.
 	ExternalEtcdNotAvailable = "ExternalEtcdNotAvailable"
+
+	// AutoscalerConstraintNotMetReason reports the Cluster status is waiting for autoscaler constraint to be met.
+	AutoscalerConstraintNotMetReason = "AutoscalerConstraintNotMet"
 )
 
 const (

pkg/clusterapi/autoscaler.go

@@ -8,9 +8,10 @@ import (
 	anywherev1 "github.com/aws/eks-anywhere/pkg/api/v1alpha1"
 )
 
+// Autoscaler annotation constants.
 const (
-	nodeGroupMinSizeAnnotation = "cluster.x-k8s.io/cluster-api-autoscaler-node-group-min-size"
-	nodeGroupMaxSizeAnnotation = "cluster.x-k8s.io/cluster-api-autoscaler-node-group-max-size"
+	NodeGroupMinSizeAnnotation = "cluster.x-k8s.io/cluster-api-autoscaler-node-group-min-size"
+	NodeGroupMaxSizeAnnotation = "cluster.x-k8s.io/cluster-api-autoscaler-node-group-max-size"
 )
 
 func ConfigureAutoscalingInMachineDeployment(md *clusterv1.MachineDeployment, autoscalingConfig *anywherev1.AutoScalingConfiguration) {
@@ -22,6 +23,6 @@ func ConfigureAutoscalingInMachineDeployment(md *clusterv1.MachineDeployment, au
 		md.ObjectMeta.Annotations = map[string]string{}
 	}
 
-	md.ObjectMeta.Annotations[nodeGroupMinSizeAnnotation] = strconv.Itoa(autoscalingConfig.MinCount)
-	md.ObjectMeta.Annotations[nodeGroupMaxSizeAnnotation] = strconv.Itoa(autoscalingConfig.MaxCount)
+	md.ObjectMeta.Annotations[NodeGroupMinSizeAnnotation] = strconv.Itoa(autoscalingConfig.MinCount)
+	md.ObjectMeta.Annotations[NodeGroupMaxSizeAnnotation] = strconv.Itoa(autoscalingConfig.MaxCount)
 }

pkg/controller/clusters/status.go

@@ -12,6 +12,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
 	anywherev1 "github.com/aws/eks-anywhere/pkg/api/v1alpha1"
+	"github.com/aws/eks-anywhere/pkg/clusterapi"
 	"github.com/aws/eks-anywhere/pkg/controller"
 )
 
@@ -198,11 +199,17 @@ func updateWorkersReadyCondition(cluster *anywherev1.Cluster, machineDeployments
 	}
 
 	totalExpected := 0
+	wngWithAutoScalingConfigurationMap := make(map[string]anywherev1.AutoScalingConfiguration)
 	for _, wng := range cluster.Spec.WorkerNodeGroupConfigurations {
-		totalExpected += *wng.Count
+		// We want to consider only the worker node groups which don't have autoscaling configuration for expected worker nodes count.
+		if wng.AutoScalingConfiguration == nil {
+			totalExpected += *wng.Count
+		} else {
+			wngWithAutoScalingConfigurationMap[wng.Name] = *wng.AutoScalingConfiguration
+		}
 	}
 
-	// First, we need aggregate the number of nodes across worker node groups  to be able to assess the condition of the workers
+	// First, we need to aggregate the number of nodes across worker node groups to be able to assess the condition of the workers
 	// as a whole.
 	totalReadyReplicas := 0
 	totalUpdatedReplicas := 0
@@ -215,6 +222,13 @@ func updateWorkersReadyCondition(cluster *anywherev1.Cluster, machineDeployments
 			return
 		}
 
+		// Skip updating the replicas for the machine deployments which have autoscaling configuration annotation
+		if md.ObjectMeta.Annotations != nil {
+			if _, ok := md.ObjectMeta.Annotations[clusterapi.NodeGroupMinSizeAnnotation]; ok {
+				continue
+			}
+		}
+
 		totalReadyReplicas += int(md.Status.ReadyReplicas)
 		totalUpdatedReplicas += int(md.Status.UpdatedReplicas)
 		totalReplicas += int(md.Status.Replicas)
@@ -253,6 +267,18 @@ func updateWorkersReadyCondition(cluster *anywherev1.Cluster, machineDeployments
 		return
 	}
 
+	for _, md := range machineDeployments {
+		if wng, exists := wngWithAutoScalingConfigurationMap[md.ObjectMeta.Name]; exists {
+			minCount := wng.MinCount
+			maxCount := wng.MaxCount
+			replicas := int(md.Status.Replicas)
+			if replicas < minCount || replicas > maxCount {
+				conditions.MarkFalse(cluster, anywherev1.WorkersReadyCondition, anywherev1.AutoscalerConstraintNotMetReason, clusterv1.ConditionSeverityInfo, "Worker nodes count for %s not between %d and %d yet (%d actual)", md.ObjectMeta.Name, minCount, maxCount, replicas)
+				return
+			}
+		}
+	}
+
 	conditions.MarkTrue(cluster, anywherev1.WorkersReadyCondition)
 }