use DeploymentAvailability to check addon health (#212)

Signed-off-by: zhujian <jiazhu@redhat.com>

go.mod

@@ -27,7 +27,7 @@ require (
 	k8s.io/klog/v2 v2.90.1
 	k8s.io/kube-aggregator v0.27.2
 	k8s.io/utils v0.0.0-20230313181309-38a27ef9d749
-	open-cluster-management.io/addon-framework v0.7.1-0.20230626092851-963716af4eed
+	open-cluster-management.io/addon-framework v0.7.1-0.20230705031704-6a328fa5cd63
 	open-cluster-management.io/api v0.11.1-0.20230703133341-6d7212c2e941
 	sigs.k8s.io/controller-runtime v0.15.0
 	sigs.k8s.io/kube-storage-version-migrator v0.0.5

go.sum

@@ -1156,8 +1156,8 @@ k8s.io/utils v0.0.0-20200324210504-a9aa75ae1b89/go.mod h1:sZAwmy6armz5eXlNoLmJcl
 k8s.io/utils v0.0.0-20201110183641-67b214c5f920/go.mod h1:jPW/WVKK9YHAvNhRxK0md/EJ228hCsBRufyofKtW8HA=
 k8s.io/utils v0.0.0-20230313181309-38a27ef9d749 h1:xMMXJlJbsU8w3V5N2FLDQ8YgU8s1EoULdbQBcAeNJkY=
 k8s.io/utils v0.0.0-20230313181309-38a27ef9d749/go.mod h1:OLgZIPagt7ERELqWJFomSt595RzquPNLL48iOWgYOg0=
-open-cluster-management.io/addon-framework v0.7.1-0.20230626092851-963716af4eed h1:fOOKf8kzVNizc5fYvMwkPy9TT/vOpojd4IIxpzh/vhw=
-open-cluster-management.io/addon-framework v0.7.1-0.20230626092851-963716af4eed/go.mod h1:Cyt5knxR+sXaKvOfUKseZDAGulS2AJz6o7a9J0WXbak=
+open-cluster-management.io/addon-framework v0.7.1-0.20230705031704-6a328fa5cd63 h1:GCsAD1jb6wqhXTHdUM/HcWzv5b2NbZ6FxpLZcxa/jhI=
+open-cluster-management.io/addon-framework v0.7.1-0.20230705031704-6a328fa5cd63/go.mod h1:V+WUFC7GD89Lc68eXSN/FJebnCH4NjrfF44VsO0YAC8=
 open-cluster-management.io/api v0.11.1-0.20230703133341-6d7212c2e941 h1:k10Sx7Th1UDyJ+GYFqWddFq+m6U7x9MHk1g8KwrYy8Y=
 open-cluster-management.io/api v0.11.1-0.20230703133341-6d7212c2e941/go.mod h1:WgKUCJ7+Bf40DsOmH1Gdkpyj3joco+QLzrlM6Ak39zE=
 rsc.io/binaryregexp v0.2.0/go.mod h1:qTv7/COck+e2FymRvadv62gMdZztPaShugOCi3I+8D8=

pkg/addon/controllers/addontemplate/controller.go

@@ -93,6 +93,7 @@ func (c *addonTemplateController) stopUnusedManagers(
 	stopFunc, ok := c.addonManagers[addOnName]
 	if ok {
 		stopFunc()
+		delete(c.addonManagers, addOnName)
 		klog.Infof("Stop the manager for addon %s", addOnName)
 	}
 }

pkg/addon/controllers/addontemplate/controller_test.go

@@ -7,10 +7,12 @@ import (
 	"time"
 
 	"github.com/openshift/library-go/pkg/operator/events/eventstesting"
+	"github.com/stretchr/testify/assert"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/client-go/dynamic/dynamicinformer"
 	dynamicfake "k8s.io/client-go/dynamic/fake"
 	fakekube "k8s.io/client-go/kubernetes/fake"
+	"k8s.io/client-go/rest"
 
 	"open-cluster-management.io/addon-framework/pkg/addonmanager/addontesting"
 	"open-cluster-management.io/addon-framework/pkg/utils"
@@ -215,3 +217,52 @@ func TestReconcile(t *testing.T) {
 		}
 	}
 }
+
+func TestRunController(t *testing.T) {
+	cases := []struct {
+		name        string
+		addonName   string
+		expectedErr string
+	}{
+		{
+			name:        "addon name empty",
+			addonName:   "",
+			expectedErr: "addon name should be set",
+		},
+		{
+			name:        "fake kubeconfig",
+			addonName:   "test",
+			expectedErr: `Get "http://localhost/api": dial tcp [::1]:80: connect: connection refused`,
+		},
+	}
+
+	for _, c := range cases {
+		fakeAddonClient := fakeaddon.NewSimpleClientset()
+		addonInformers := addoninformers.NewSharedInformerFactory(fakeAddonClient, 10*time.Minute)
+		fakeDynamicClient := dynamicfake.NewSimpleDynamicClient(runtime.NewScheme())
+		dynamicInformerFactory := dynamicinformer.NewDynamicSharedInformerFactory(fakeDynamicClient, 0)
+		fakeClusterClient := fakecluster.NewSimpleClientset()
+		clusterInformers := clusterv1informers.NewSharedInformerFactory(fakeClusterClient, 10*time.Minute)
+		fakeWorkClient := fakework.NewSimpleClientset()
+		workInformers := workinformers.NewSharedInformerFactory(fakeWorkClient, 10*time.Minute)
+		hubKubeClient := fakekube.NewSimpleClientset()
+		controller := &addonTemplateController{
+			kubeConfig:       &rest.Config{},
+			kubeClient:       hubKubeClient,
+			addonClient:      fakeAddonClient,
+			cmaLister:        addonInformers.Addon().V1alpha1().ClusterManagementAddOns().Lister(),
+			addonManagers:    make(map[string]context.CancelFunc),
+			addonInformers:   addonInformers,
+			clusterInformers: clusterInformers,
+			dynamicInformers: dynamicInformerFactory,
+			workInformers:    workInformers,
+		}
+		ctx := context.TODO()
+
+		err := controller.runController(ctx, c.addonName)
+		if len(c.expectedErr) == 0 {
+			assert.NoError(t, err)
+		}
+		assert.EqualErrorf(t, err, c.expectedErr, "name : %s, expected error %v, but got %v", c.name, c.expectedErr, err)
+	}
+}

pkg/addon/templateagent/template_agent.go

@@ -4,7 +4,6 @@ import (
 	"fmt"
 
 	"github.com/valyala/fasttemplate"
-	appsv1 "k8s.io/api/apps/v1"
 	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/runtime/schema"
@@ -105,9 +104,11 @@ func (a *CRDTemplateAgentAddon) GetAgentAddonOptions() agent.AgentAddonOptions {
 		supportedConfigGVRs = append(supportedConfigGVRs, gvr)
 	}
 	return agent.AgentAddonOptions{
-		AddonName:           a.addonName,
-		InstallStrategy:     nil,
-		HealthProber:        nil,
+		AddonName:       a.addonName,
+		InstallStrategy: nil,
+		HealthProber: &agent.HealthProber{
+			Type: agent.HealthProberTypeDeploymentAvailability,
+		},
 		SupportedConfigGVRs: supportedConfigGVRs,
 		Registration: &agent.RegistrationOption{
 			CSRConfigurations: a.TemplateCSRConfigurationsFunc(),
@@ -159,7 +160,7 @@ func (a *CRDTemplateAgentAddon) decorateObjects(
 		newImageDecorator(privateValues),
 	}
 	for index, obj := range objects {
-		deployment, err := a.convertToDeployment(obj)
+		deployment, err := utils.ConvertToDeployment(obj)
 		if err != nil {
 			continue
 		}
@@ -176,26 +177,6 @@ func (a *CRDTemplateAgentAddon) decorateObjects(
 	return objects, nil
 }
 
-func (a *CRDTemplateAgentAddon) convertToDeployment(obj runtime.Object) (*appsv1.Deployment, error) {
-	if obj.GetObjectKind().GroupVersionKind().Group != "apps" ||
-		obj.GetObjectKind().GroupVersionKind().Kind != "Deployment" {
-		return nil, fmt.Errorf("not deployment object, %v", obj.GetObjectKind())
-	}
-
-	deployment := &appsv1.Deployment{}
-	uobj, ok := obj.(*unstructured.Unstructured)
-	if !ok {
-		return deployment, fmt.Errorf("not unstructured object, %v", obj.GetObjectKind())
-	}
-
-	err := runtime.DefaultUnstructuredConverter.
-		FromUnstructured(uobj.Object, deployment)
-	if err != nil {
-		return nil, err
-	}
-	return deployment, nil
-}
-
 // GetDesiredAddOnTemplateByAddon returns the desired template of the addon
 func (a *CRDTemplateAgentAddon) GetDesiredAddOnTemplateByAddon(
 	addon *addonapiv1alpha1.ManagedClusterAddOn) (*addonapiv1alpha1.AddOnTemplate, error) {

test/e2e/addonmanagement_test.go

@@ -169,6 +169,10 @@ var _ = ginkgo.Describe("Enable addon management feature gate", ginkgo.Label("ad
 		ginkgo.By(fmt.Sprintf("clean klusterlet %v resources after the test case", klusterletName))
 		gomega.Expect(t.cleanKlusterletResources(klusterletName, clusterName)).To(gomega.BeNil())
 
+		ginkgo.By(fmt.Sprintf("Cleaning managed cluster namespace %s", clusterName))
+		err = t.HubKubeClient.CoreV1().Namespaces().Delete(context.TODO(), clusterName, metav1.DeleteOptions{})
+		gomega.Expect(err).ToNot(gomega.HaveOccurred())
+
 		// disable addon management feature gate
 		gomega.Eventually(func() error {
 			clusterManager, err := t.OperatorClient.OperatorV1().ClusterManagers().Get(context.TODO(), "cluster-manager", metav1.GetOptions{})

vendor/modules.txt

@@ -1406,7 +1406,7 @@ k8s.io/utils/path
 k8s.io/utils/pointer
 k8s.io/utils/strings/slices
 k8s.io/utils/trace
-# open-cluster-management.io/addon-framework v0.7.1-0.20230626092851-963716af4eed
+# open-cluster-management.io/addon-framework v0.7.1-0.20230705031704-6a328fa5cd63
 ## explicit; go 1.19
 open-cluster-management.io/addon-framework/pkg/addonfactory
 open-cluster-management.io/addon-framework/pkg/addonmanager