Make it possible to create clusters with only karpenter-managed worker nodes

Apply suggestions from code review

Co-authored-by: John Gardiner Myers <jgmyers@proofpoint.com>

cmd/kops/create_cluster.go

@@ -35,6 +35,9 @@ import (
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/util/sets"
 	"k8s.io/klog/v2"
+	"k8s.io/kubectl/pkg/util/i18n"
+	"k8s.io/kubectl/pkg/util/templates"
+
 	kopsbase "k8s.io/kops"
 	"k8s.io/kops/cmd/kops/util"
 	api "k8s.io/kops/pkg/apis/kops"
@@ -53,8 +56,6 @@ import (
 	"k8s.io/kops/upup/pkg/fi"
 	"k8s.io/kops/upup/pkg/fi/cloudup"
 	"k8s.io/kops/upup/pkg/fi/utils"
-	"k8s.io/kubectl/pkg/util/i18n"
-	"k8s.io/kubectl/pkg/util/templates"
 )
 
 type CreateClusterOptions struct {
@@ -449,6 +450,12 @@ func NewCmdCreateCluster(f *util.Factory, out io.Writer) *cobra.Command {
 		return pflag.NormalizedName(name)
 	})
 
+	if featureflag.Karpenter.Enabled() {
+		cmd.Flags().StringVar(&options.InstanceManager, "instance-manager", options.InstanceManager, "Instance manager to use (cloudgroups or karpenter. Default: cloudgroups)")
+		cmd.RegisterFlagCompletionFunc("instance-manager", func(cmd *cobra.Command, args []string, toComplete string) ([]string, cobra.ShellCompDirective) {
+			return []string{"cloudgroups", "karpenter"}, cobra.ShellCompDirectiveNoFileComp
+		})
+	}
 	return cmd
 }
 

tests/integration/update_cluster/karpenter/data/aws_s3_bucket_object_minimal.example.com-addons-bootstrap_content

@@ -54,7 +54,7 @@ spec:
     version: 9.99.0
   - id: k8s-1.19
     manifest: karpenter.sh/k8s-1.19.yaml
-    manifestHash: 45610f3413287dd87de888b7dbbe31931a10f7125aec6ee88516d0d92c9494b4
+    manifestHash: 632dd9ea482692ffcaf421dafc79081b9f0a745b154948ee5558c47ed1802d52
     name: karpenter.sh
     selector:
       k8s-addon: karpenter.sh

tests/integration/update_cluster/karpenter/data/aws_s3_bucket_object_minimal.example.com-addons-karpenter.sh-k8s-1.19_content

@@ -653,10 +653,15 @@ spec:
           name: token-amazonaws-com
           readOnly: true
       dnsPolicy: Default
+      nodeSelector:
+        node-role.kubernetes.io/master: ""
       priorityClassName: system-cluster-critical
       securityContext:
         fsGroup: 10001
       serviceAccountName: karpenter
+      tolerations:
+      - key: node-role.kubernetes.io/master
+        operator: Exists
       volumes:
       - name: token-amazonaws-com
         projected:
@@ -744,10 +749,19 @@ spec:
         - mountPath: /var/run/secrets/amazonaws.com/
           name: token-amazonaws-com
           readOnly: true
+      nodeSelector:
+        node-role.kubernetes.io/master: ""
       priorityClassName: system-cluster-critical
       securityContext:
         fsGroup: 10001
       serviceAccountName: karpenter
+      tolerations:
+      - key: node.cloudprovider.kubernetes.io/uninitialized
+        operator: Exists
+      - key: node.kubernetes.io/not-ready
+        operator: Exists
+      - key: node-role.kubernetes.io/master
+        operator: Exists
       volumes:
       - name: token-amazonaws-com
         projected:

upup/models/cloudup/resources/addons/karpenter.sh/k8s-1.19.yaml.template

@@ -488,6 +488,12 @@ spec:
                 - linux
               - key: karpenter.sh/provisioner-name
                 operator: DoesNotExist
+      tolerations:
+      - key: node-role.kubernetes.io/master
+        operator: Exists
+      nodeSelector:
+        node-role.kubernetes.io/master: ""
+
 ---
 # Source: karpenter/templates/webhook/deployment.yaml
 apiVersion: apps/v1
@@ -554,6 +560,15 @@ spec:
                 - linux
               - key: karpenter.sh/provisioner-name
                 operator: DoesNotExist
+      tolerations:
+      - key: node.cloudprovider.kubernetes.io/uninitialized
+        operator: Exists
+      - key: node.kubernetes.io/not-ready
+        operator: Exists
+      - key: node-role.kubernetes.io/master
+        operator: Exists
+      nodeSelector:
+        node-role.kubernetes.io/master: ""
 ---
 # Source: karpenter/templates/webhook/webhooks.yaml
 apiVersion: admissionregistration.k8s.io/v1

upup/pkg/fi/cloudup/new_cluster.go

@@ -27,6 +27,7 @@ import (
 	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/sets"
 	"k8s.io/klog/v2"
+
 	"k8s.io/kops"
 	api "k8s.io/kops/pkg/apis/kops"
 	"k8s.io/kops/pkg/apis/kops/model"
@@ -146,6 +147,9 @@ type NewClusterOptions struct {
 	// APISSLCertificate is the SSL certificate to use for the API loadbalancer.
 	// Currently only supported in AWS.
 	APISSLCertificate string
+
+	// InstanceManager specifies which manager to use for managing instances.
+	InstanceManager string
 }
 
 func (o *NewClusterOptions) InitDefaults() {
@@ -155,6 +159,7 @@ func (o *NewClusterOptions) InitDefaults() {
 	o.Networking = "kubenet"
 	o.Topology = api.TopologyPublic
 	o.DNSType = string(api.DNSTypePublic)
+	o.InstanceManager = "cloudgroups"
 }
 
 type NewClusterResult struct {
@@ -286,9 +291,27 @@ func NewCluster(opt *NewClusterOptions, clientset simple.Clientset) (*NewCluster
 		return nil, err
 	}
 
-	nodes, err := setupNodes(opt, &cluster, zoneToSubnetMap)
+	var nodes []*api.InstanceGroup
-	if err != nil {
+
-		return nil, err
+	switch opt.InstanceManager {
+	case "karpenter":
+		if opt.DiscoveryStore == "" {
+			return nil, fmt.Errorf("karpenter requires --discovery-store")
+		}
+		cluster.Spec.Karpenter = &api.KarpenterConfig{
+			Enabled: true,
+		}
+		nodes, err = setupKarpenterNodes(opt, &cluster, zoneToSubnetMap)
+		if err != nil {
+			return nil, err
+		}
+	case "cloudgroups":
+		nodes, err = setupNodes(opt, &cluster, zoneToSubnetMap)
+		if err != nil {
+			return nil, err
+		}
+	default:
+		return nil, fmt.Errorf("invalid value %q for --instance-manager", opt.InstanceManager)
 	}
 
 	apiservers, err := setupAPIServers(opt, &cluster, zoneToSubnetMap)
@@ -842,6 +865,20 @@ func setupNodes(opt *NewClusterOptions, cluster *api.Cluster, zoneToSubnetMap ma
 	return nodes, nil
 }
 
+func setupKarpenterNodes(opt *NewClusterOptions, cluster *api.Cluster, zoneToSubnetMap map[string]*api.ClusterSubnetSpec) ([]*api.InstanceGroup, error) {
+	g := &api.InstanceGroup{}
+	g.Spec.Role = api.InstanceGroupRoleNode
+	g.Spec.Manager = api.InstanceManagerKarpenter
+	g.ObjectMeta.Name = "nodes"
+
+	g.Spec.InstanceMetadata = &api.InstanceMetadataOptions{
+		HTTPPutResponseHopLimit: fi.Int64(1),
+		HTTPTokens:              fi.String("required"),
+	}
+
+	return []*api.InstanceGroup{g}, nil
+}
+
 func setupAPIServers(opt *NewClusterOptions, cluster *api.Cluster, zoneToSubnetMap map[string]*api.ClusterSubnetSpec) ([]*api.InstanceGroup, error) {
 	cloudProvider := api.CloudProviderID(cluster.Spec.CloudProvider)