add opt-in field for disabling instance refresh

Author: harveyxia

cmd/clusterawsadm/credentials/credentials.go

@@ -30,6 +30,7 @@ import (
 
 // AWSCredentialsTemplate generates an AWS credentials file that can
 // be loaded by the various SDKs.
+//
 //nolint:gosec
 const AWSCredentialsTemplate = `[default]
 aws_access_key_id = {{ .AccessKeyID }}

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

@@ -789,6 +789,11 @@ spec:
                 description: RefreshPreferences describes set of preferences associated
                   with the instance refresh request.
                 properties:
+                  disable:
+                    description: Disable, if true, disables instance refresh from
+                      triggering when new launch templates are detected. This is useful
+                      in scenarios where ASG nodes are externally managed.
+                    type: boolean
                   instanceWarmup:
                     description: The number of seconds until a newly launched instance
                       is configured and ready to use. During this time, the next replacement

exp/api/v1beta1/conversion.go

@@ -41,19 +41,22 @@ 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 {
+		dst.Spec.RefreshPreferences.Disable = restored.Spec.RefreshPreferences.Disable
+	}
 
 	return nil
 }
 
 // ConvertFrom converts the v1beta2 AWSMachinePool receiver to v1beta1 AWSMachinePool.
-func (r *AWSMachinePool) ConvertFrom(srcRaw conversion.Hub) error {
+func (dst *AWSMachinePool) ConvertFrom(srcRaw conversion.Hub) error {
 	src := srcRaw.(*infrav1exp.AWSMachinePool)
 
-	if err := Convert_v1beta2_AWSMachinePool_To_v1beta1_AWSMachinePool(src, r, nil); err != nil {
+	if err := Convert_v1beta2_AWSMachinePool_To_v1beta1_AWSMachinePool(src, dst, nil); err != nil {
 		return err
 	}
 
-	return utilconversion.MarshalData(src, r)
+	return utilconversion.MarshalData(src, dst)
 }
 
 // ConvertTo converts the v1beta1 AWSMachinePoolList receiver to a v1beta2 AWSMachinePoolList.
@@ -175,3 +178,9 @@ func Convert_v1beta2_AutoScalingGroup_To_v1beta1_AutoScalingGroup(in *infrav1exp
 	// explicitly ignore CurrentlySuspended.
 	return autoConvert_v1beta2_AutoScalingGroup_To_v1beta1_AutoScalingGroup(in, out, s)
 }
+
+// Convert_v1beta2_RefreshPreferences_To_v1beta1_RefreshPreferences converts the v1beta2 RefreshPreferences receiver to a v1beta1 RefreshPreferences.
+func Convert_v1beta2_RefreshPreferences_To_v1beta1_RefreshPreferences(in *infrav1exp.RefreshPreferences, out *RefreshPreferences, s apiconversion.Scope) error {
+	// spec.refreshPreferences.disable has been added to v1beta2.
+	return autoConvert_v1beta2_RefreshPreferences_To_v1beta1_RefreshPreferences(in, out, s)
+}

exp/api/v1beta1/zz_generated.conversion.go

@@ -139,6 +139,11 @@ func (s *SuspendProcessesTypes) ConvertSetValuesToStringSlice() []string {
 
 // RefreshPreferences defines the specs for instance refreshing.
 type RefreshPreferences struct {
+	// Disable, if true, disables instance refresh from triggering when new launch templates are detected.
+	// This is useful in scenarios where ASG nodes are externally managed.
+	// +optional
+	Disable bool `json:"disable,omitempty"`
+
 	// The strategy to use for the instance refresh. The only valid value is Rolling.
 	// A rolling update is an update that is applied to all instances in an Auto
 	// Scaling group until all instances have been updated.

exp/api/v1beta2/awsmachinepool_types.go

@@ -232,6 +232,11 @@ func (r *AWSMachinePoolReconciler) reconcileNormal(ctx context.Context, machineP
 		return asgsvc.CanStartASGInstanceRefresh(machinePoolScope)
 	}
 	runPostLaunchTemplateUpdateOperation := func() error {
+		// skip instance refresh if explicitly disabled
+		if machinePoolScope.AWSMachinePool.Spec.RefreshPreferences != nil && machinePoolScope.AWSMachinePool.Spec.RefreshPreferences.Disable {
+			machinePoolScope.V(4).Info("instance refresh disabled, skipping instance refresh")
+			return nil
+		}
 		// After creating a new version of launch template, instance refresh is required
 		// to trigger a rolling replacement of all previously launched instances.
 		// If ONLY the userdata changed, previously launched instances continue to use the old launch