Support setting maxHealthyPercentage to configure ASG instance refresh

config/crd/bases/infrastructure.cluster.x-k8s.io_awsmachinepools.yaml

@@ -987,6 +987,22 @@ spec:
                       The default is to use the value for the health check grace period defined for the group.
                     format: int64
                     type: integer
+                  maxHealthyPercentage:
+                    description: |-
+                      The amount of capacity as a percentage in ASG that can be in service and healthy, or pending,
+                      to support your workload when replacing instances.
+                      The value is expressed as a percentage of the desired capacity of the ASG. Value range is 100 to 200.
+                      If you specify MaxHealthyPercentage , you must also specify MinHealthyPercentage , and the difference between
+                      them cannot be greater than 100.
+
+
+                      A larger range increases the number of instances that can be replaced at the same time.
+
+
+                      If you do not specify this property, the default is 100 percent, or the percentage set in the instance
+                      maintenance policy for the Auto Scaling group, if defined.
+                    format: int64
+                    type: integer
                   minHealthyPercentage:
                     description: |-
                       The amount of capacity as a percentage in ASG that must remain healthy

exp/api/v1beta1/conversion.go

@@ -41,8 +41,9 @@ func (src *AWSMachinePool) ConvertTo(dstRaw conversion.Hub) error {
 	if restored.Spec.SuspendProcesses != nil {
 		dst.Spec.SuspendProcesses = restored.Spec.SuspendProcesses
 	}
-	if dst.Spec.RefreshPreferences != nil && restored.Spec.RefreshPreferences != nil {
+	if restored.Spec.RefreshPreferences != nil {
 		dst.Spec.RefreshPreferences.Disable = restored.Spec.RefreshPreferences.Disable
+		dst.Spec.RefreshPreferences.MaxHealthyPercentage = restored.Spec.RefreshPreferences.MaxHealthyPercentage
 	}
 	if restored.Spec.AWSLaunchTemplate.InstanceMetadataOptions != nil {
 		dst.Spec.AWSLaunchTemplate.InstanceMetadataOptions = restored.Spec.AWSLaunchTemplate.InstanceMetadataOptions

exp/api/v1beta2/awsmachinepool_types.go

@@ -176,6 +176,19 @@ type RefreshPreferences struct {
 	// during an instance refresh. The default is 90.
 	// +optional
 	MinHealthyPercentage *int64 `json:"minHealthyPercentage,omitempty"`
+
+	// The amount of capacity as a percentage in ASG that can be in service and healthy, or pending,
+	// to support your workload when replacing instances.
+	// The value is expressed as a percentage of the desired capacity of the ASG. Value range is 100 to 200.
+	// If you specify MaxHealthyPercentage , you must also specify MinHealthyPercentage , and the difference between
+	// them cannot be greater than 100.
+	//
+	// A larger range increases the number of instances that can be replaced at the same time.
+	//
+	// If you do not specify this property, the default is 100 percent, or the percentage set in the instance
+	// maintenance policy for the Auto Scaling group, if defined.
+	// +optional
+	MaxHealthyPercentage *int64 `json:"maxHealthyPercentage,omitempty"`
 }
 
 // AWSMachinePoolStatus defines the observed state of AWSMachinePool.

exp/api/v1beta2/awsmachinepool_webhook.go

@@ -109,6 +109,7 @@ func (r *AWSMachinePool) validateAdditionalSecurityGroups() field.ErrorList {
 	}
 	return allErrs
 }
+
 func (r *AWSMachinePool) validateSpotInstances() field.ErrorList {
 	var allErrs field.ErrorList
 	if r.Spec.AWSLaunchTemplate.SpotMarketOptions != nil && r.Spec.MixedInstancesPolicy != nil {
@@ -133,6 +134,26 @@ func (r *AWSMachinePool) validateIgnition() field.ErrorList {
 	return allErrs
 }
 
+func (r *AWSMachinePool) validateRefreshPreferences() field.ErrorList {
+	var allErrs field.ErrorList
+
+	if r.Spec.RefreshPreferences == nil {
+		return allErrs
+	}
+
+	if r.Spec.RefreshPreferences.MaxHealthyPercentage != nil && r.Spec.RefreshPreferences.MinHealthyPercentage == nil {
+		allErrs = append(allErrs, field.Forbidden(field.NewPath("spec.refreshPreferences.maxHealthyPercentage"), "If you specify spec.refreshPreferences.maxHealthyPercentage, you must also specify spec.refreshPreferences.minHealthyPercentage"))
+	}
+
+	if r.Spec.RefreshPreferences.MaxHealthyPercentage != nil && r.Spec.RefreshPreferences.MinHealthyPercentage != nil {
+		if *r.Spec.RefreshPreferences.MaxHealthyPercentage-*r.Spec.RefreshPreferences.MinHealthyPercentage > 100 {
+			allErrs = append(allErrs, field.Forbidden(field.NewPath("spec.refreshPreferences.maxHealthyPercentage"), "the difference between spec.refreshPreferences.maxHealthyPercentage and spec.refreshPreferences.minHealthyPercentage cannot be greater than 100"))
+		}
+	}
+
+	return allErrs
+}
+
 // ValidateCreate will do any extra validation when creating a AWSMachinePool.
 func (r *AWSMachinePool) ValidateCreate() (admission.Warnings, error) {
 	log.Info("AWSMachinePool validate create", "machine-pool", klog.KObj(r))
@@ -146,6 +167,7 @@ func (r *AWSMachinePool) ValidateCreate() (admission.Warnings, error) {
 	allErrs = append(allErrs, r.validateAdditionalSecurityGroups()...)
 	allErrs = append(allErrs, r.validateSpotInstances()...)
 	allErrs = append(allErrs, r.validateIgnition()...)
+	allErrs = append(allErrs, r.validateRefreshPreferences()...)
 
 	if len(allErrs) == 0 {
 		return nil, nil
@@ -167,6 +189,7 @@ func (r *AWSMachinePool) ValidateUpdate(_ runtime.Object) (admission.Warnings, e
 	allErrs = append(allErrs, r.validateSubnets()...)
 	allErrs = append(allErrs, r.validateAdditionalSecurityGroups()...)
 	allErrs = append(allErrs, r.validateSpotInstances()...)
+	allErrs = append(allErrs, r.validateRefreshPreferences()...)
 
 	if len(allErrs) == 0 {
 		return nil, nil

exp/api/v1beta2/awsmachinepool_webhook_test.go

@@ -153,6 +153,27 @@ func TestAWSMachinePoolValidateCreate(t *testing.T) {
 			},
 			wantErr: true,
 		},
+		{
+			name: "Should fail if MaxHealthyPercentage is set, but MinHealthyPercentage is not set",
+			pool: &AWSMachinePool{
+				Spec: AWSMachinePoolSpec{
+					RefreshPreferences: &RefreshPreferences{MaxHealthyPercentage: aws.Int64(100)},
+				},
+			},
+			wantErr: true,
+		},
+		{
+			name: "Should fail if the difference between MaxHealthyPercentage and MinHealthyPercentage is greater than 100",
+			pool: &AWSMachinePool{
+				Spec: AWSMachinePoolSpec{
+					RefreshPreferences: &RefreshPreferences{
+						MaxHealthyPercentage: aws.Int64(150),
+						MinHealthyPercentage: aws.Int64(25),
+					},
+				},
+			},
+			wantErr: true,
+		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -287,6 +308,27 @@ func TestAWSMachinePoolValidateUpdate(t *testing.T) {
 			},
 			wantErr: true,
 		},
+		{
+			name: "Should fail if MaxHealthyPercentage is set, but MinHealthyPercentage is not set",
+			new: &AWSMachinePool{
+				Spec: AWSMachinePoolSpec{
+					RefreshPreferences: &RefreshPreferences{MaxHealthyPercentage: aws.Int64(100)},
+				},
+			},
+			wantErr: true,
+		},
+		{
+			name: "Should fail if the difference between MaxHealthyPercentage and MinHealthyPercentage is greater than 100",
+			new: &AWSMachinePool{
+				Spec: AWSMachinePoolSpec{
+					RefreshPreferences: &RefreshPreferences{
+						MaxHealthyPercentage: aws.Int64(150),
+						MinHealthyPercentage: aws.Int64(25),
+					},
+				},
+			},
+			wantErr: true,
+		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {

pkg/cloud/services/autoscaling/autoscalinggroup.go

@@ -373,7 +373,7 @@ func (s *Service) CancelASGInstanceRefresh(scope *scope.MachinePoolScope) error
 // StartASGInstanceRefresh will start an ASG instance refresh.
 func (s *Service) StartASGInstanceRefresh(scope *scope.MachinePoolScope) error {
 	strategy := ptr.To[string](autoscaling.RefreshStrategyRolling)
-	var minHealthyPercentage, instanceWarmup *int64
+	var minHealthyPercentage, maxHealthyPercentage, instanceWarmup *int64
 	if scope.AWSMachinePool.Spec.RefreshPreferences != nil {
 		if scope.AWSMachinePool.Spec.RefreshPreferences.Strategy != nil {
 			strategy = scope.AWSMachinePool.Spec.RefreshPreferences.Strategy
@@ -384,6 +384,9 @@ func (s *Service) StartASGInstanceRefresh(scope *scope.MachinePoolScope) error {
 		if scope.AWSMachinePool.Spec.RefreshPreferences.MinHealthyPercentage != nil {
 			minHealthyPercentage = scope.AWSMachinePool.Spec.RefreshPreferences.MinHealthyPercentage
 		}
+		if scope.AWSMachinePool.Spec.RefreshPreferences.MaxHealthyPercentage != nil {
+			maxHealthyPercentage = scope.AWSMachinePool.Spec.RefreshPreferences.MaxHealthyPercentage
+		}
 	}
 
 	input := &autoscaling.StartInstanceRefreshInput{
@@ -392,6 +395,7 @@ func (s *Service) StartASGInstanceRefresh(scope *scope.MachinePoolScope) error {
 		Preferences: &autoscaling.RefreshPreferences{
 			InstanceWarmup:       instanceWarmup,
 			MinHealthyPercentage: minHealthyPercentage,
+			MaxHealthyPercentage: maxHealthyPercentage,
 		},
 	}
 

pkg/cloud/services/autoscaling/autoscalinggroup_test.go

@@ -1207,6 +1207,7 @@ func TestServiceStartASGInstanceRefresh(t *testing.T) {
 					Preferences: &autoscaling.RefreshPreferences{
 						InstanceWarmup:       aws.Int64(100),
 						MinHealthyPercentage: aws.Int64(80),
+						MaxHealthyPercentage: aws.Int64(100),
 					},
 				})).
 					Return(nil, awserrors.NewNotFound("not found"))
@@ -1222,6 +1223,7 @@ func TestServiceStartASGInstanceRefresh(t *testing.T) {
 					Preferences: &autoscaling.RefreshPreferences{
 						InstanceWarmup:       aws.Int64(100),
 						MinHealthyPercentage: aws.Int64(80),
+						MaxHealthyPercentage: aws.Int64(100),
 					},
 				})).
 					Return(&autoscaling.StartInstanceRefreshOutput{}, nil)
@@ -1316,6 +1318,7 @@ func getMachinePoolScope(client client.Client, clusterScope *scope.ClusterScope)
 				Strategy:             aws.String("Rolling"),
 				InstanceWarmup:       aws.Int64(100),
 				MinHealthyPercentage: aws.Int64(80),
+				MaxHealthyPercentage: aws.Int64(100),
 			},
 			MixedInstancesPolicy: &expinfrav1.MixedInstancesPolicy{
 				InstancesDistribution: &expinfrav1.InstancesDistribution{