Merge pull request #576 from giantswarm/cherry-pick-bugfixes

Cherry-pick bugfixes into v2.3.x release branch

api/v1beta2/tags.go

@@ -190,6 +190,11 @@ const (
 
 	// MachineNameTagKey is the key for machine name.
 	MachineNameTagKey = "MachineName"
+
+	// LaunchTemplateBootstrapDataSecret is the tag we use to store the `<namespace>/<name>`
+	// of the bootstrap secret that was used to create the user data for the latest launch
+	// template version.
+	LaunchTemplateBootstrapDataSecret = NameAWSProviderPrefix + "bootstrap-data-secret"
 )
 
 // ClusterTagKey generates the key for resources associated with a cluster.

exp/controllers/awsmachinepool_controller.go

@@ -60,6 +60,7 @@ type AWSMachinePoolReconciler struct {
 	WatchFilterValue             string
 	asgServiceFactory            func(cloud.ClusterScoper) services.ASGInterface
 	ec2ServiceFactory            func(scope.EC2Scope) services.EC2Interface
+	reconcileServiceFactory      func(scope.EC2Scope) services.MachinePoolReconcileInterface
 	TagUnmanagedNetworkResources bool
 }
 
@@ -78,6 +79,14 @@ func (r *AWSMachinePoolReconciler) getEC2Service(scope scope.EC2Scope) services.
 	return ec2.NewService(scope)
 }
 
+func (r *AWSMachinePoolReconciler) getReconcileService(scope scope.EC2Scope) services.MachinePoolReconcileInterface {
+	if r.reconcileServiceFactory != nil {
+		return r.reconcileServiceFactory(scope)
+	}
+
+	return ec2.NewService(scope)
+}
+
 // +kubebuilder:rbac:groups=infrastructure.cluster.x-k8s.io,resources=awsmachinepools,verbs=get;list;watch;update;patch;delete
 // +kubebuilder:rbac:groups=infrastructure.cluster.x-k8s.io,resources=awsmachinepools/status,verbs=get;update;patch
 // +kubebuilder:rbac:groups=cluster.x-k8s.io,resources=machinepools;machinepools/status,verbs=get;list;watch;patch
@@ -225,14 +234,32 @@ func (r *AWSMachinePoolReconciler) reconcileNormal(ctx context.Context, machineP
 
 	ec2Svc := r.getEC2Service(ec2Scope)
 	asgsvc := r.getASGService(clusterScope)
+	reconSvc := r.getReconcileService(ec2Scope)
+
+	// Find existing ASG
+	asg, err := r.findASG(machinePoolScope, asgsvc)
+	if err != nil {
+		conditions.MarkUnknown(machinePoolScope.AWSMachinePool, expinfrav1.ASGReadyCondition, expinfrav1.ASGNotFoundReason, err.Error())
+		return err
+	}
 
 	canUpdateLaunchTemplate := func() (bool, error) {
 		// If there is a change: before changing the template, check if there exist an ongoing instance refresh,
 		// because only 1 instance refresh can be "InProgress". If template is updated when refresh cannot be started,
 		// that change will not trigger a refresh. Do not start an instance refresh if only userdata changed.
+		if asg == nil {
+			// If the ASG hasn't been created yet, there is no need to check if we can start the instance refresh.
+			// But we want to update the LaunchTemplate because an error in the LaunchTemplate may be blocking the ASG creation.
+			return true, nil
+		}
 		return asgsvc.CanStartASGInstanceRefresh(machinePoolScope)
 	}
 	runPostLaunchTemplateUpdateOperation := func() error {
+		// skip instance refresh if ASG is not created yet
+		if asg == nil {
+			machinePoolScope.Debug("ASG does not exist yet, skipping instance refresh")
+			return nil
+		}
 		// skip instance refresh if explicitly disabled
 		if machinePoolScope.AWSMachinePool.Spec.RefreshPreferences != nil && machinePoolScope.AWSMachinePool.Spec.RefreshPreferences.Disable {
 			machinePoolScope.Debug("instance refresh disabled, skipping instance refresh")
@@ -250,7 +277,7 @@ func (r *AWSMachinePoolReconciler) reconcileNormal(ctx context.Context, machineP
 		machinePoolScope.Info("starting instance refresh", "number of instances", machinePoolScope.MachinePool.Spec.Replicas)
 		return asgsvc.StartASGInstanceRefresh(machinePoolScope)
 	}
-	if err := ec2Svc.ReconcileLaunchTemplate(machinePoolScope, canUpdateLaunchTemplate, runPostLaunchTemplateUpdateOperation); err != nil {
+	if err := reconSvc.ReconcileLaunchTemplate(machinePoolScope, ec2Svc, canUpdateLaunchTemplate, runPostLaunchTemplateUpdateOperation); err != nil {
 		r.Recorder.Eventf(machinePoolScope.AWSMachinePool, corev1.EventTypeWarning, "FailedLaunchTemplateReconcile", "Failed to reconcile launch template: %v", err)
 		machinePoolScope.Error(err, "failed to reconcile launch template")
 		return err
@@ -259,13 +286,6 @@ func (r *AWSMachinePoolReconciler) reconcileNormal(ctx context.Context, machineP
 	// set the LaunchTemplateReady condition
 	conditions.MarkTrue(machinePoolScope.AWSMachinePool, expinfrav1.LaunchTemplateReadyCondition)
 
-	// Find existing ASG
-	asg, err := r.findASG(machinePoolScope, asgsvc)
-	if err != nil {
-		conditions.MarkUnknown(machinePoolScope.AWSMachinePool, expinfrav1.ASGReadyCondition, expinfrav1.ASGNotFoundReason, err.Error())
-		return err
-	}
-
 	if asg == nil {
 		// Create new ASG
 		if err := r.createPool(machinePoolScope, clusterScope); err != nil {
@@ -305,7 +325,7 @@ func (r *AWSMachinePoolReconciler) reconcileNormal(ctx context.Context, machineP
 			ResourceService: asgsvc,
 		},
 	}
-	err = ec2Svc.ReconcileTags(machinePoolScope, resourceServiceToUpdate)
+	err = reconSvc.ReconcileTags(machinePoolScope, resourceServiceToUpdate)
 	if err != nil {
 		return errors.Wrap(err, "error updating tags")
 	}
@@ -366,7 +386,7 @@ func (r *AWSMachinePoolReconciler) reconcileDelete(machinePoolScope *scope.Machi
 	}
 
 	launchTemplateID := machinePoolScope.AWSMachinePool.Status.LaunchTemplateID
-	launchTemplate, _, err := ec2Svc.GetLaunchTemplate(machinePoolScope.LaunchTemplateName())
+	launchTemplate, _, _, err := ec2Svc.GetLaunchTemplate(machinePoolScope.LaunchTemplateName())
 	if err != nil {
 		return err
 	}
@@ -410,7 +430,7 @@ func (r *AWSMachinePoolReconciler) updatePool(machinePoolScope *scope.MachinePoo
 
 	asgDiff := diffASG(machinePoolScope, existingASG)
 	if asgDiff != "" {
-		machinePoolScope.Debug("asg diff detected", "diff", subnetDiff)
+		machinePoolScope.Debug("asg diff detected", "asgDiff", asgDiff, "subnetDiff", subnetDiff)
 	}
 	if asgDiff != "" || subnetDiff != "" {
 		machinePoolScope.Info("updating AutoScalingGroup")