topology-updater: Add owner reference to NRT object

Signed-off-by: Oleg Zhurakivskyy <[email protected]>

Author: ozhuraki

deployment/base/rbac-topologyupdater/topologyupdater-clusterrole.yaml

@@ -10,6 +10,12 @@ rules:
   verbs:
   - get
   - list
+- apiGroups:
+  - ""
+  resources:
+  - namespaces
+  verbs:
+  - get
 - apiGroups:
   - ""
   resources:

deployment/helm/node-feature-discovery/templates/clusterrole.yaml

@@ -58,6 +58,12 @@ rules:
   verbs:
   - get
   - list
+- apiGroups:
+  - ""
+  resources:
+  - namespaces
+  verbs:
+  - get
 - apiGroups:
     - ""
   resources:

pkg/nfd-topology-updater/nfd-topology-updater.go

@@ -26,6 +26,7 @@ import (
 
 	"k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
 	k8sclient "k8s.io/client-go/kubernetes"
 	"k8s.io/klog/v2"
 	kubeletconfigv1beta1 "k8s.io/kubelet/config/v1beta1"
@@ -80,6 +81,9 @@ type nfdTopologyUpdater struct {
 	eventSource         <-chan kubeletnotifier.Info
 	configFilePath      string
 	config              *NFDConfig
+	kubernetesNamespace string
+	ownerRefs           []metav1.OwnerReference
+	k8sClient           k8sclient.Interface
 	kubeletConfigFunc   func() (*kubeletconfigv1beta1.KubeletConfiguration, error)
 }
 
@@ -105,6 +109,8 @@ func NewTopologyUpdater(args Args, resourcemonitorArgs resourcemonitor.Args) (Nf
 		nodeName:            utils.NodeName(),
 		eventSource:         eventSource,
 		config:              &NFDConfig{},
+		kubernetesNamespace: utils.GetKubernetesNamespace(),
+		ownerRefs:           []metav1.OwnerReference{},
 		kubeletConfigFunc:   kubeletConfigFunc,
 	}
 	if args.ConfigFile != "" {
@@ -146,6 +152,7 @@ func (w *nfdTopologyUpdater) Run() error {
 	if err != nil {
 		return err
 	}
+	w.k8sClient = k8sClient
 
 	if err := w.configure(); err != nil {
 		return fmt.Errorf("faild to configure Node Feature Discovery Topology Updater: %w", err)
@@ -222,11 +229,29 @@ func (w *nfdTopologyUpdater) Stop() {
 }
 
 func (w *nfdTopologyUpdater) updateNodeResourceTopology(zoneInfo v1alpha2.ZoneList, scanResponse resourcemonitor.ScanResponse, readKubeletConfig bool) error {
+
+	if len(w.ownerRefs) == 0 {
+		ns, err := w.k8sClient.CoreV1().Namespaces().Get(context.TODO(), w.kubernetesNamespace, metav1.GetOptions{})
+		if err != nil {
+			klog.ErrorS(err, "Cannot get NodeResourceTopology owner reference")
+		} else {
+			w.ownerRefs = []metav1.OwnerReference{
+				{
+					APIVersion: ns.APIVersion,
+					Kind:       ns.Kind,
+					Name:       ns.Name,
+					UID:        types.UID(ns.UID),
+				},
+			}
+		}
+	}
+
 	nrt, err := w.topoClient.TopologyV1alpha2().NodeResourceTopologies().Get(context.TODO(), w.nodeName, metav1.GetOptions{})
 	if errors.IsNotFound(err) {
 		nrtNew := v1alpha2.NodeResourceTopology{
 			ObjectMeta: metav1.ObjectMeta{
-				Name: w.nodeName,
+				Name:            w.nodeName,
+				OwnerReferences: w.ownerRefs,
 			},
 			Zones:      zoneInfo,
 			Attributes: v1alpha2.AttributeList{},
@@ -248,6 +273,7 @@ func (w *nfdTopologyUpdater) updateNodeResourceTopology(zoneInfo v1alpha2.ZoneLi
 
 	nrtMutated := nrt.DeepCopy()
 	nrtMutated.Zones = zoneInfo
+	nrtMutated.OwnerReferences = w.ownerRefs
 
 	attributes := scanResponse.Attributes
 

test/e2e/topology_updater_test.go

@@ -261,6 +261,44 @@ var _ = NFDDescribe(Label("nfd-topology-updater"), func() {
 				return true
 			}, time.Minute, 5*time.Second).Should(BeTrue(), "didn't get updated node topology info")
 		})
+
+		It("should check that that topology object is garbage colleted", func(ctx context.Context) {
+
+			By("Check if the topology object has owner reference")
+			ns, err := f.ClientSet.CoreV1().Namespaces().Get(ctx, f.Namespace.Name, metav1.GetOptions{})
+			Expect(err).NotTo(HaveOccurred())
+
+			t, err := topologyClient.TopologyV1alpha2().NodeResourceTopologies().Get(ctx, topologyUpdaterNode.Name, metav1.GetOptions{})
+			Expect(err).NotTo(HaveOccurred())
+
+			owned := false
+			for _, r := range t.OwnerReferences {
+				if r.UID == ns.UID {
+					owned = true
+					break
+				}
+			}
+			Expect(owned).Should(BeTrue())
+
+			By("Deleting the nfd-topology namespace")
+			err = f.ClientSet.CoreV1().Namespaces().Delete(ctx, f.Namespace.Name, metav1.DeleteOptions{})
+			Expect(err).NotTo(HaveOccurred())
+
+			By("checking that topology was garbage collected")
+			Eventually(func() bool {
+				t, err := topologyClient.TopologyV1alpha2().NodeResourceTopologies().Get(ctx, topologyUpdaterNode.Name, metav1.GetOptions{})
+				if err != nil {
+					if apierrors.IsNotFound(err) {
+						framework.Logf("missing node topology resource for %q", topologyUpdaterNode.Name)
+						return false
+					}
+					framework.Logf("failed to get the node topology resource: %v", err)
+					return true
+				}
+				framework.Logf("topology resource: %v", t)
+				return true
+			}).WithPolling(15 * time.Second).WithTimeout(60 * time.Second).Should(BeFalse())
+		})
 	})
 
 	When("sleep interval disabled", func() {

test/e2e/utils/rbac.go

@@ -22,6 +22,7 @@ import (
 
 	corev1 "k8s.io/api/core/v1"
 	rbacv1 "k8s.io/api/rbac/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	clientset "k8s.io/client-go/kubernetes"
 )
@@ -105,10 +106,6 @@ func DeconfigureRBAC(ctx context.Context, cs clientset.Interface, ns string) err
 	if err != nil {
 		return err
 	}
-	err = cs.RbacV1().RoleBindings(ns).Delete(ctx, "nfd-worker-e2e", metav1.DeleteOptions{})
-	if err != nil {
-		return err
-	}
 	err = cs.RbacV1().ClusterRoleBindings().Delete(ctx, "nfd-gc-e2e", metav1.DeleteOptions{})
 	if err != nil {
 		return err
@@ -121,11 +118,24 @@ func DeconfigureRBAC(ctx context.Context, cs clientset.Interface, ns string) err
 	if err != nil {
 		return err
 	}
-	err = cs.RbacV1().Roles(ns).Delete(ctx, "nfd-worker-e2e", metav1.DeleteOptions{})
+	err = cs.RbacV1().ClusterRoles().Delete(ctx, "nfd-gc-e2e", metav1.DeleteOptions{})
 	if err != nil {
 		return err
 	}
-	err = cs.RbacV1().ClusterRoles().Delete(ctx, "nfd-gc-e2e", metav1.DeleteOptions{})
+
+	// Skip the deletion of namespaced objects in case the namespace is gone
+	_, err = cs.CoreV1().Namespaces().Get(ctx, ns, metav1.GetOptions{})
+	if err != nil {
+		if apierrors.IsNotFound(err) {
+			return nil
+		}
+		return err
+	}
+	err = cs.RbacV1().RoleBindings(ns).Delete(ctx, "nfd-worker-e2e", metav1.DeleteOptions{})
+	if err != nil {
+		return err
+	}
+	err = cs.RbacV1().Roles(ns).Delete(ctx, "nfd-worker-e2e", metav1.DeleteOptions{})
 	if err != nil {
 		return err
 	}
@@ -251,6 +261,11 @@ func createClusterRoleTopologyUpdater(ctx context.Context, cs clientset.Interfac
 				Resources: []string{"pods"},
 				Verbs:     []string{"get", "list", "watch"},
 			},
+			{
+				APIGroups: []string{""},
+				Resources: []string{"namespaces"},
+				Verbs:     []string{"get"},
+			},
 			{
 				APIGroups: []string{"topology.node.k8s.io"},
 				Resources: []string{"noderesourcetopologies"},