Fix: removing detached clusters from resource trackers (#6728)

* removing detached clusters from resource trackers

Signed-off-by: Pushparaj Shetty KS <kspushparajshetty@gmail.com>
Signed-off-by: Pushparaj Shetty K S <kspushparajshetty@gmail.com>

* resolve merge conflicts

Signed-off-by: Pushparaj Shetty K S <kspushparajshetty@gmail.com>

* resolved code conflicts

Signed-off-by: Chaitanya Reddy Onteddu <chaitanyareddy0702@gmail.com>
Signed-off-by: Pushparaj Shetty K S <kspushparajshetty@gmail.com>

* update TestGetAddonStatus test case

Signed-off-by: Chaitanya Reddy Onteddu <chaitanyareddy0702@gmail.com>

---------

Signed-off-by: Pushparaj Shetty KS <kspushparajshetty@gmail.com>
Signed-off-by: Pushparaj Shetty K S <kspushparajshetty@gmail.com>
Signed-off-by: Chaitanya Reddy Onteddu <chaitanyareddy0702@gmail.com>
Co-authored-by: Shivin Gopalani <sgopalani@guidewire.com>
Co-authored-by: Pushparaj Shetty K S <kspushparajshetty@gmail.com>

pkg/addon/addon_test.go

@@ -378,7 +378,8 @@ func TestGetAddonStatus(t *testing.T) {
 	})
 
 	cli := test.MockClient{
-		MockGet: getFunc,
+		MockGet:  getFunc,
+		MockList: listFunc,
 	}
 
 	cases := []struct {
@@ -410,6 +411,10 @@ func TestGetAddonStatus(t *testing.T) {
 	}
 }
 
+func listFunc(ctx context.Context, list client.ObjectList, opts ...client.ListOption) error {
+	return nil
+}
+
 func TestGetAddonVersionMeetSystemRequirement(t *testing.T) {
 	server := httptest.NewServer(helmHandler)
 	defer server.Close()

pkg/addon/helper.go

@@ -123,6 +123,14 @@ func EnableAddonByLocalDir(ctx context.Context, name string, dir string, cli cli
 // GetAddonStatus is general func for cli and apiServer get addon status
 func GetAddonStatus(ctx context.Context, cli client.Client, name string) (Status, error) {
 	var addonStatus Status
+	joinedClusters, err := multicluster.NewClusterClient(cli).List(ctx)
+	if err != nil {
+		return addonStatus, errors.Wrap(err, "failed to list registered clusters")
+	}
+	var joinedClusterMap = make(map[string]bool)
+	for _, joinedCluster := range joinedClusters.Items {
+		joinedClusterMap[joinedCluster.Name] = true
+	}
 
 	app, err := FetchAddonRelatedApp(ctx, cli, name)
 	if err != nil {
@@ -143,8 +151,12 @@ func GetAddonStatus(ctx context.Context, cli client.Client, name string) (Status
 			r.Cluster = multicluster.ClusterLocalName
 		}
 		// TODO(wonderflow): we should collect all the necessary information as observability, currently we only collect cluster name
-		clusters[r.Cluster] = make(map[string]interface{})
+		// If cluster is not registered in KubeVela then skip it.
+		if joinedClusterMap[r.Cluster] {
+			clusters[r.Cluster] = make(map[string]interface{})
+		}
 	}
+
 	addonStatus.Clusters = clusters
 
 	if app.Status.Workflow != nil && app.Status.Workflow.Suspend {

pkg/multicluster/cluster_management.go

@@ -44,6 +44,7 @@ import (
 	ocmclusterv1 "open-cluster-management.io/api/cluster/v1"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
+	"github.com/oam-dev/kubevela/apis/core.oam.dev/v1beta1"
 	"github.com/oam-dev/kubevela/pkg/utils"
 	cmdutil "github.com/oam-dev/kubevela/pkg/utils/util"
 )
@@ -486,6 +487,9 @@ func (op DetachClusterManagedClusterKubeConfigPathOption) ApplyToArgs(args *Deta
 
 // DetachCluster detach cluster by name, if cluster is using by application, it will return error
 func DetachCluster(ctx context.Context, cli client.Client, clusterName string, options ...DetachClusterOption) error {
+	if err := removeClusterFromResourceTrackers(ctx, cli, clusterName); err != nil {
+		return fmt.Errorf("error in removing cluster references from resourcetrackers: %w", err)
+	}
 	args := newDetachClusterArgs(options...)
 	if clusterName == ClusterLocalName {
 		return ErrReservedLocalClusterName
@@ -612,6 +616,30 @@ func getMutableClusterSecret(ctx context.Context, c client.Client, clusterName s
 	return clusterSecret, nil
 }
 
+// removeClusterFromResourceTrackers removes cluster references from all resource trackers.
+func removeClusterFromResourceTrackers(ctx context.Context, cli client.Client, clusterName string) error {
+	rts := v1beta1.ResourceTrackerList{}
+	if err := cli.List(ctx, &rts); err != nil {
+		return fmt.Errorf("unable to list resourcetrackers due to error: %w", err)
+	}
+	for i := range rts.Items {
+		managedResources := rts.Items[i].Spec.ManagedResources
+		var result []v1beta1.ManagedResource
+		for _, mr := range managedResources {
+			if mr.ClusterObjectReference.Cluster != clusterName {
+				result = append(result, mr)
+			}
+		}
+		if len(rts.Items[i].Spec.ManagedResources) != len(result) {
+			rts.Items[i].Spec.ManagedResources = result
+			if err := cli.Update(ctx, &rts.Items[i]); err != nil {
+				return fmt.Errorf("error in updating resourcetracker %s: %w", rts.Items[i].Name, err)
+			}
+		}
+	}
+	return nil
+}
+
 func getTokenFromExec(execConfig *clientcmdapi.ExecConfig) (string, error) {
 	// #nosec G204 -- This is intentionally running an exec command with user-provided input
 	// The execConfig comes from the kubeconfig which should be trusted in this context