chore: Drop karpenter.k8s.aws/cluster tag from launch templates (#7379

)

pkg/apis/v1/labels.go

@@ -126,7 +126,6 @@ var (
 	AnnotationEC2NodeClassHashVersion         = apis.Group + "/ec2nodeclass-hash-version"
 	AnnotationInstanceTagged                  = apis.Group + "/tagged"
 
-	TagNodeClaim             = coreapis.Group + "/nodeclaim"
-	TagManagedLaunchTemplate = apis.Group + "/cluster"
-	TagName                  = "Name"
+	TagNodeClaim = coreapis.Group + "/nodeclaim"
+	TagName      = "Name"
 )

pkg/cloudprovider/cloudprovider.go

@@ -35,6 +35,7 @@ import (
 
 	"github.com/aws/karpenter-provider-aws/pkg/apis"
 	v1 "github.com/aws/karpenter-provider-aws/pkg/apis/v1"
+	"github.com/aws/karpenter-provider-aws/pkg/operator/options"
 	"github.com/aws/karpenter-provider-aws/pkg/utils"
 
 	"github.com/samber/lo"
@@ -102,7 +103,7 @@ func (c *CloudProvider) Create(ctx context.Context, nodeClaim *karpv1.NodeClaim)
 	if len(instanceTypes) == 0 {
 		return nil, cloudprovider.NewInsufficientCapacityError(fmt.Errorf("all requested instance types were unavailable during launch"))
 	}
-	instance, err := c.instanceProvider.Create(ctx, nodeClass, nodeClaim, instanceTypes)
+	instance, err := c.instanceProvider.Create(ctx, nodeClass, nodeClaim, getTags(ctx, nodeClass, nodeClaim), instanceTypes)
 	if err != nil {
 		return nil, fmt.Errorf("creating instance, %w", err)
 	}
@@ -227,6 +228,16 @@ func (c *CloudProvider) GetSupportedNodeClasses() []status.Object {
 	return []status.Object{&v1.EC2NodeClass{}}
 }
 
+func getTags(ctx context.Context, nodeClass *v1.EC2NodeClass, nodeClaim *karpv1.NodeClaim) map[string]string {
+	staticTags := map[string]string{
+		fmt.Sprintf("kubernetes.io/cluster/%s", options.FromContext(ctx).ClusterName): "owned",
+		karpv1.NodePoolLabelKey: nodeClaim.Labels[karpv1.NodePoolLabelKey],
+		v1.EKSClusterNameTagKey: options.FromContext(ctx).ClusterName,
+		v1.LabelNodeClass:       nodeClass.Name,
+	}
+	return lo.Assign(nodeClass.Spec.Tags, staticTags)
+}
+
 func (c *CloudProvider) resolveNodeClassFromNodeClaim(ctx context.Context, nodeClaim *karpv1.NodeClaim) (*v1.EC2NodeClass, error) {
 	nodeClass := &v1.EC2NodeClass{}
 	if err := c.kubeClient.Get(ctx, types.NamespacedName{Name: nodeClaim.Spec.NodeClassRef.Name}, nodeClass); err != nil {

pkg/controllers/nodeclass/termination/suite_test.go

@@ -139,9 +139,9 @@ var _ = Describe("NodeClass Termination", func() {
 	})
 	It("should succeed to delete the launch template", func() {
 		ltName1 := aws.String(fake.LaunchTemplateName())
-		awsEnv.EC2API.LaunchTemplates.Store(ltName1, ec2types.LaunchTemplate{LaunchTemplateName: ltName1, LaunchTemplateId: aws.String(fake.LaunchTemplateID()), Tags: []ec2types.Tag{{Key: aws.String("karpenter.k8s.aws/cluster"), Value: aws.String("test-cluster")}, {Key: aws.String("karpenter.k8s.aws/ec2nodeclass"), Value: aws.String(nodeClass.Name)}}})
+		awsEnv.EC2API.LaunchTemplates.Store(ltName1, ec2types.LaunchTemplate{LaunchTemplateName: ltName1, LaunchTemplateId: aws.String(fake.LaunchTemplateID()), Tags: []ec2types.Tag{{Key: aws.String("eks:eks-cluster-name"), Value: aws.String("test-cluster")}, {Key: aws.String("karpenter.k8s.aws/ec2nodeclass"), Value: aws.String(nodeClass.Name)}}})
 		ltName2 := aws.String(fake.LaunchTemplateName())
-		awsEnv.EC2API.LaunchTemplates.Store(ltName2, ec2types.LaunchTemplate{LaunchTemplateName: ltName2, LaunchTemplateId: aws.String(fake.LaunchTemplateID()), Tags: []ec2types.Tag{{Key: aws.String("karpenter.k8s.aws/cluster"), Value: aws.String("test-cluster")}, {Key: aws.String("karpenter.k8s.aws/ec2nodeclass"), Value: aws.String(nodeClass.Name)}}})
+		awsEnv.EC2API.LaunchTemplates.Store(ltName2, ec2types.LaunchTemplate{LaunchTemplateName: ltName2, LaunchTemplateId: aws.String(fake.LaunchTemplateID()), Tags: []ec2types.Tag{{Key: aws.String("eks:eks-cluster-name"), Value: aws.String("test-cluster")}, {Key: aws.String("karpenter.k8s.aws/ec2nodeclass"), Value: aws.String(nodeClass.Name)}}})
 		_, ok := awsEnv.EC2API.LaunchTemplates.Load(ltName1)
 		Expect(ok).To(BeTrue())
 		_, ok = awsEnv.EC2API.LaunchTemplates.Load(ltName2)

pkg/fake/utils.go

@@ -139,7 +139,7 @@ func Filter(filters []ec2types.Filter, id, name string, tags []ec2types.Tag) boo
 // nolint: gocyclo
 func matchTags(tags []ec2types.Tag, filter ec2types.Filter) bool {
 	if strings.HasPrefix(*filter.Name, "tag:") {
-		tagKey := strings.Split(*filter.Name, ":")[1]
+		_, tagKey, _ := strings.Cut(*filter.Name, ":")
 		for _, val := range filter.Values {
 			for _, tag := range tags {
 				if tagKey == *tag.Key && (val == "*" || val == *tag.Value) {

pkg/providers/instance/instance.go

@@ -69,7 +69,7 @@ var (
 )
 
 type Provider interface {
-	Create(context.Context, *v1.EC2NodeClass, *karpv1.NodeClaim, []*cloudprovider.InstanceType) (*Instance, error)
+	Create(context.Context, *v1.EC2NodeClass, *karpv1.NodeClaim, map[string]string, []*cloudprovider.InstanceType) (*Instance, error)
 	Get(context.Context, string) (*Instance, error)
 	List(context.Context) ([]*Instance, error)
 	Delete(context.Context, string) error
@@ -97,7 +97,7 @@ func NewDefaultProvider(ctx context.Context, region string, ec2api sdk.EC2API, u
 	}
 }
 
-func (p *DefaultProvider) Create(ctx context.Context, nodeClass *v1.EC2NodeClass, nodeClaim *karpv1.NodeClaim, instanceTypes []*cloudprovider.InstanceType) (*Instance, error) {
+func (p *DefaultProvider) Create(ctx context.Context, nodeClass *v1.EC2NodeClass, nodeClaim *karpv1.NodeClaim, tags map[string]string, instanceTypes []*cloudprovider.InstanceType) (*Instance, error) {
 	schedulingRequirements := scheduling.NewNodeSelectorRequirementsWithMinValues(nodeClaim.Spec.Requirements...)
 	// Only filter the instances if there are no minValues in the requirement.
 	if !schedulingRequirements.HasMinValues() {
@@ -108,7 +108,6 @@ func (p *DefaultProvider) Create(ctx context.Context, nodeClass *v1.EC2NodeClass
 		log.FromContext(ctx).Error(err, "truncating instance types")
 		return nil, fmt.Errorf("truncating instance types, %w", err)
 	}
-	tags := getTags(ctx, nodeClass, nodeClaim)
 	fleetInstance, err := p.launchInstance(ctx, nodeClass, nodeClaim, instanceTypes, tags)
 	if awserrors.IsLaunchTemplateNotFound(err) {
 		// retry once if launch template is not found. This allows karpenter to generate a new LT if the
@@ -268,16 +267,6 @@ func (p *DefaultProvider) launchInstance(ctx context.Context, nodeClass *v1.EC2N
 	return createFleetOutput.Instances[0], nil
 }
 
-func getTags(ctx context.Context, nodeClass *v1.EC2NodeClass, nodeClaim *karpv1.NodeClaim) map[string]string {
-	staticTags := map[string]string{
-		fmt.Sprintf("kubernetes.io/cluster/%s", options.FromContext(ctx).ClusterName): "owned",
-		karpv1.NodePoolLabelKey: nodeClaim.Labels[karpv1.NodePoolLabelKey],
-		v1.EKSClusterNameTagKey: options.FromContext(ctx).ClusterName,
-		v1.LabelNodeClass:       nodeClass.Name,
-	}
-	return lo.Assign(nodeClass.Spec.Tags, staticTags)
-}
-
 func (p *DefaultProvider) checkODFallback(nodeClaim *karpv1.NodeClaim, instanceTypes []*cloudprovider.InstanceType, launchTemplateConfigs []ec2types.FleetLaunchTemplateConfigRequest) error {
 	// only evaluate for on-demand fallback if the capacity type for the request is OD and both OD and spot are allowed in requirements
 	if p.getCapacityType(nodeClaim, instanceTypes) != karpv1.CapacityTypeOnDemand || !scheduling.NewNodeSelectorRequirementsWithMinValues(nodeClaim.Spec.Requirements...).Get(karpv1.CapacityTypeLabelKey).Has(karpv1.CapacityTypeSpot) {

pkg/providers/instance/suite_test.go

@@ -133,7 +133,7 @@ var _ = Describe("InstanceProvider", func() {
 		instanceTypes = lo.Filter(instanceTypes, func(i *corecloudprovider.InstanceType, _ int) bool { return i.Name == "m5.xlarge" })
 
 		// Since all the capacity pools are ICEd. This should return back an ICE error
-		instance, err := awsEnv.InstanceProvider.Create(ctx, nodeClass, nodeClaim, instanceTypes)
+		instance, err := awsEnv.InstanceProvider.Create(ctx, nodeClass, nodeClaim, nil, instanceTypes)
 		Expect(corecloudprovider.IsInsufficientCapacityError(err)).To(BeTrue())
 		Expect(instance).To(BeNil())
 	})

pkg/providers/launchtemplate/launchtemplate.go

@@ -261,7 +261,7 @@ func (p *DefaultProvider) createLaunchTemplate(ctx context.Context, options *ami
 		TagSpecifications: []ec2types.TagSpecification{
 			{
 				ResourceType: ec2types.ResourceTypeLaunchTemplate,
-				Tags:         utils.MergeTags(options.Tags, map[string]string{v1.TagManagedLaunchTemplate: options.ClusterName, v1.LabelNodeClass: options.NodeClassName}),
+				Tags:         utils.MergeTags(options.Tags),
 			},
 		},
 	})
@@ -348,12 +348,12 @@ func (p *DefaultProvider) volumeSize(quantity *resource.Quantity) *int32 {
 // Any error during hydration will result in a panic
 func (p *DefaultProvider) hydrateCache(ctx context.Context) {
 	clusterName := options.FromContext(ctx).ClusterName
-	ctx = log.IntoContext(ctx, log.FromContext(ctx).WithValues("tag-key", v1.TagManagedLaunchTemplate, "tag-value", clusterName))
+	ctx = log.IntoContext(ctx, log.FromContext(ctx).WithValues("tag-key", v1.EKSClusterNameTagKey, "tag-value", clusterName))
 
 	paginator := ec2.NewDescribeLaunchTemplatesPaginator(p.ec2api, &ec2.DescribeLaunchTemplatesInput{
 		Filters: []ec2types.Filter{
 			{
-				Name:   aws.String(fmt.Sprintf("tag:%s", v1.TagManagedLaunchTemplate)),
+				Name:   aws.String(fmt.Sprintf("tag:%s", v1.EKSClusterNameTagKey)),
 				Values: []string{clusterName},
 			},
 		},
@@ -400,7 +400,7 @@ func (p *DefaultProvider) DeleteAll(ctx context.Context, nodeClass *v1.EC2NodeCl
 	paginator := ec2.NewDescribeLaunchTemplatesPaginator(p.ec2api, &ec2.DescribeLaunchTemplatesInput{
 		Filters: []ec2types.Filter{
 			{
-				Name:   aws.String(fmt.Sprintf("tag:%s", v1.TagManagedLaunchTemplate)),
+				Name:   aws.String(fmt.Sprintf("tag:%s", v1.EKSClusterNameTagKey)),
 				Values: []string{clusterName},
 			},
 			{

website/content/en/preview/upgrading/upgrade-guide.md

@@ -38,6 +38,7 @@ WHEN CREATING A NEW SECTION OF THE UPGRADE GUIDANCE FOR NEWER VERSIONS, ENSURE T
 * Bottlerocket AMIFamily now supports `instanceStorePolicy: RAID0`. This means that Karpenter will auto-generate userData to RAID0 your instance store volumes (similar to AL2 and AL2023) when specifying this value.
   * Note: This userData configuration is _only_ valid on Bottlerocket v1.22.0+. If you are using an earlier version of a Bottlerocket image (< v1.22.0) with `amiFamily: Bottlerocket` and `instanceStorePolicy: RAID0`, nodes will fail to join the cluster.  
 * The AWS Neuron accelerator well known name label (`karpenter.k8s.aws/instance-accelerator-name`) values now reflect their correct names of `trainium`, `inferentia`, and `inferentia2`. Previously, all Neuron accelerators were assigned the label name of `inferentia`.
+* Karpenter drops the internal `karpenter.k8s.aws/cluster` tag used for launch template management in favor of `eks:eks-cluster-name` and consistency with other Karpenter-provisioned resources
 
 ### Upgrading to `1.0.0`+