chore: removing finalizers and controllerswitch (#3779)

Signed-off-by: Jaydip Gabani <[email protected]>

config/rbac/role.yaml

@@ -168,15 +168,6 @@ rules:
   - patch
   - update
   - watch
-- apiGroups:
-  - templates.gatekeeper.sh
-  resources:
-  - constrainttemplates/finalizers
-  verbs:
-  - delete
-  - get
-  - patch
-  - update
 - apiGroups:
   - templates.gatekeeper.sh
   resources:

main.go

@@ -284,10 +284,6 @@ func innerMain() int {
 		close(setupFinished)
 	}
 
-	// ControllerSwitch will be used to disable controllers during our teardown process,
-	// avoiding conflicts in finalizer cleanup.
-	sw := watch.NewSwitch()
-
 	// Setup tracker and register readiness probe.
 	tracker, err := readiness.SetupTracker(mgr, mutation.Enabled(), *externaldata.ExternalDataEnabled, *expansion.ExpansionEnabled)
 	if err != nil {
@@ -316,7 +312,7 @@ func innerMain() int {
 	setupErr := make(chan error)
 	ctx := ctrl.SetupSignalHandler()
 	go func() {
-		setupErr <- setupControllers(ctx, mgr, sw, tracker, setupFinished)
+		setupErr <- setupControllers(ctx, mgr, tracker, setupFinished)
 	}()
 
 	setupLog.Info("starting manager")
@@ -348,20 +344,15 @@ blockingLoop:
 			break blockingLoop
 		}
 	}
-
-	// Manager stops controllers asynchronously.
-	// Instead, we use ControllerSwitch to synchronously prevent them from doing more work.
-	// This can be removed when finalizer and status teardown is removed.
 	setupLog.Info("disabling controllers...")
-	sw.Stop()
 
 	if hadError {
 		return 1
 	}
 	return 0
 }
 
-func setupControllers(ctx context.Context, mgr ctrl.Manager, sw *watch.ControllerSwitch, tracker *readiness.Tracker, setupFinished chan struct{}) error {
+func setupControllers(ctx context.Context, mgr ctrl.Manager, tracker *readiness.Tracker, setupFinished chan struct{}) error {
 	// Block until the setup (certificate generation) finishes.
 	<-setupFinished
 
@@ -508,17 +499,16 @@ func setupControllers(ctx context.Context, mgr ctrl.Manager, sw *watch.Controlle
 	}
 
 	opts := controller.Dependencies{
-		CFClient:         client,
-		WatchManger:      wm,
-		SyncEventsCh:     events,
-		CacheMgr:         cm,
-		ControllerSwitch: sw,
-		Tracker:          tracker,
-		ProcessExcluder:  processExcluder,
-		MutationSystem:   mutationSystem,
-		ExpansionSystem:  expansionSystem,
-		ProviderCache:    providerCache,
-		PubsubSystem:     pubsubSystem,
+		CFClient:        client,
+		WatchManger:     wm,
+		SyncEventsCh:    events,
+		CacheMgr:        cm,
+		Tracker:         tracker,
+		ProcessExcluder: processExcluder,
+		MutationSystem:  mutationSystem,
+		ExpansionSystem: expansionSystem,
+		ProviderCache:   providerCache,
+		PubsubSystem:    pubsubSystem,
 	}
 
 	if err := controller.AddToManager(mgr, &opts); err != nil {

manifest_staging/charts/gatekeeper/templates/gatekeeper-manager-role-clusterrole.yaml

@@ -177,15 +177,6 @@ rules:
   - patch
   - update
   - watch
-- apiGroups:
-  - templates.gatekeeper.sh
-  resources:
-  - constrainttemplates/finalizers
-  verbs:
-  - delete
-  - get
-  - patch
-  - update
 - apiGroups:
   - templates.gatekeeper.sh
   resources:

manifest_staging/deploy/gatekeeper.yaml

@@ -4967,15 +4967,6 @@ rules:
   - patch
   - update
   - watch
-- apiGroups:
-  - templates.gatekeeper.sh
-  resources:
-  - constrainttemplates/finalizers
-  verbs:
-  - delete
-  - get
-  - patch
-  - update
 - apiGroups:
   - templates.gatekeeper.sh
   resources:

pkg/controller/config/config_controller.go

@@ -28,7 +28,6 @@ import (
 	"github.com/open-policy-agent/gatekeeper/v3/pkg/keys"
 	"github.com/open-policy-agent/gatekeeper/v3/pkg/readiness"
 	"github.com/open-policy-agent/gatekeeper/v3/pkg/util"
-	"github.com/open-policy-agent/gatekeeper/v3/pkg/watch"
 	corev1 "k8s.io/api/core/v1"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -53,28 +52,23 @@ var (
 )
 
 type Adder struct {
-	ControllerSwitch *watch.ControllerSwitch
-	Tracker          *readiness.Tracker
-	CacheManager     *cm.CacheManager
+	Tracker      *readiness.Tracker
+	CacheManager *cm.CacheManager
 	// GetPod returns an instance of the currently running Gatekeeper pod
 	GetPod func(context.Context) (*corev1.Pod, error)
 }
 
 // Add creates a new ConfigController and adds it to the Manager with default RBAC. The Manager will set fields on the Controller
 // and Start it when the Manager is Started.
 func (a *Adder) Add(mgr manager.Manager) error {
-	r, err := newReconciler(mgr, a.CacheManager, a.ControllerSwitch, a.Tracker, a.GetPod)
+	r, err := newReconciler(mgr, a.CacheManager, a.Tracker, a.GetPod)
 	if err != nil {
 		return err
 	}
 
 	return add(mgr, r)
 }
 
-func (a *Adder) InjectControllerSwitch(cs *watch.ControllerSwitch) {
-	a.ControllerSwitch = cs
-}
-
 func (a *Adder) InjectTracker(t *readiness.Tracker) {
 	a.Tracker = t
 }
@@ -88,7 +82,7 @@ func (a *Adder) InjectGetPod(getPod func(ctx context.Context) (*corev1.Pod, erro
 }
 
 // newReconciler returns a new reconcile.Reconciler.
-func newReconciler(mgr manager.Manager, cm *cm.CacheManager, cs *watch.ControllerSwitch, tracker *readiness.Tracker, getPod func(context.Context) (*corev1.Pod, error)) (*ReconcileConfig, error) {
+func newReconciler(mgr manager.Manager, cm *cm.CacheManager, tracker *readiness.Tracker, getPod func(context.Context) (*corev1.Pod, error)) (*ReconcileConfig, error) {
 	if cm == nil {
 		return nil, fmt.Errorf("cacheManager must be non-nil")
 	}
@@ -98,7 +92,6 @@ func newReconciler(mgr manager.Manager, cm *cm.CacheManager, cs *watch.Controlle
 		writer:       mgr.GetClient(),
 		statusClient: mgr.GetClient(),
 		scheme:       mgr.GetScheme(),
-		cs:           cs,
 		cacheManager: cm,
 		tracker:      tracker,
 		getPod:       getPod,
@@ -138,7 +131,6 @@ type ReconcileConfig struct {
 
 	scheme       *runtime.Scheme
 	cacheManager *cm.CacheManager
-	cs           *watch.ControllerSwitch
 
 	tracker *readiness.Tracker
 
@@ -155,15 +147,6 @@ type ReconcileConfig struct {
 // and what is in the Config.Spec
 // Automatically generate RBAC rules to allow the Controller to read all things (for sync).
 func (r *ReconcileConfig) Reconcile(ctx context.Context, request reconcile.Request) (reconcile.Result, error) {
-	// Short-circuit if shutting down.
-	if r.cs != nil {
-		running := r.cs.Enter()
-		defer r.cs.Exit()
-		if !running {
-			return reconcile.Result{}, nil
-		}
-	}
-
 	// Fetch the Config instance
 	if request.NamespacedName != keys.Config {
 		log.Info("Ignoring unsupported config name", "namespace", request.NamespacedName.Namespace, "name", request.NamespacedName.Name)

pkg/controller/config/config_controller_test.go

@@ -125,7 +125,6 @@ func TestReconcile(t *testing.T) {
 
 	dataClient := &fakes.FakeCfClient{}
 
-	cs := watch.NewSwitch()
 	tracker, err := readiness.SetupTracker(mgr, false, false, false)
 	if err != nil {
 		t.Fatal(err)
@@ -158,7 +157,7 @@ func TestReconcile(t *testing.T) {
 		fakes.WithName("no-pod"),
 	)
 
-	rec, err := newReconciler(mgr, cacheManager, cs, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
+	rec, err := newReconciler(mgr, cacheManager, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
 	require.NoError(t, err)
 
 	// Wrap the Controller Reconcile function so it writes each request to a map when it is finished reconciling.
@@ -290,8 +289,6 @@ func TestReconcile(t *testing.T) {
 
 	// fooPod should be namespace excluded, hence not added to the cache
 	require.False(t, dataClient.Contains(map[fakes.CfDataKey]interface{}{{Gvk: fooPod.GroupVersionKind(), Key: "default"}: struct{}{}}))
-
-	cs.Stop()
 }
 
 // tests that expectations for sync only resource gets canceled when it gets deleted.
@@ -424,7 +421,6 @@ func setupController(ctx context.Context, mgr manager.Manager, wm *watch.Manager
 		}
 	}
 
-	cs := watch.NewSwitch()
 	processExcluder := process.Get()
 	syncMetricsCache := syncutil.NewMetricsCache()
 	reg, err := wm.NewRegistrar(
@@ -453,7 +449,7 @@ func setupController(ctx context.Context, mgr manager.Manager, wm *watch.Manager
 		fakes.WithName("no-pod"),
 	)
 
-	rec, err := newReconciler(mgr, cacheManager, cs, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
+	rec, err := newReconciler(mgr, cacheManager, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
 	if err != nil {
 		return nil, fmt.Errorf("creating reconciler: %w", err)
 	}
@@ -600,7 +596,6 @@ func TestConfig_Retries(t *testing.T) {
 	c := testclient.NewRetryClient(mgr.GetClient())
 
 	dataClient := &fakes.FakeCfClient{}
-	cs := watch.NewSwitch()
 	tracker, err := readiness.SetupTracker(mgr, false, false, false)
 	if err != nil {
 		t.Fatal(err)
@@ -632,7 +627,7 @@ func TestConfig_Retries(t *testing.T) {
 		fakes.WithName("no-pod"),
 	)
 
-	rec, _ := newReconciler(mgr, cacheManager, cs, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
+	rec, _ := newReconciler(mgr, cacheManager, tracker, func(context.Context) (*v1.Pod, error) { return pod, nil })
 	err = add(mgr, rec)
 	if err != nil {
 		t.Fatal(err)

pkg/controller/configstatus/configstatus_controller.go

@@ -46,8 +46,6 @@ type Adder struct {
 	WatchManager *watch.Manager
 }
 
-func (a *Adder) InjectControllerSwitch(_ *watch.ControllerSwitch) {}
-
 func (a *Adder) InjectTracker(_ *readiness.Tracker) {}
 
 // Add creates a new config Status Controller and adds it to the Manager. The Manager will set fields on the Controller

pkg/controller/constraint/constraint_controller.go

@@ -87,7 +87,6 @@ type Adder struct {
 	CFClient         *constraintclient.Client
 	ConstraintsCache *ConstraintsCache
 	WatchManager     *watch.Manager
-	ControllerSwitch *watch.ControllerSwitch
 	Events           <-chan event.GenericEvent
 	Tracker          *readiness.Tracker
 	GetPod           func(context.Context) (*corev1.Pod, error)
@@ -107,10 +106,6 @@ func (a *Adder) InjectWatchManager(w *watch.Manager) {
 	a.WatchManager = w
 }
 
-func (a *Adder) InjectControllerSwitch(cs *watch.ControllerSwitch) {
-	a.ControllerSwitch = cs
-}
-
 func (a *Adder) InjectTracker(t *readiness.Tracker) {
 	a.Tracker = t
 }
@@ -127,7 +122,7 @@ func (a *Adder) Add(mgr manager.Manager) error {
 		return err
 	}
 
-	r := newReconciler(mgr, a.CFClient, a.ControllerSwitch, reporter, a.ConstraintsCache, a.Tracker)
+	r := newReconciler(mgr, a.CFClient, reporter, a.ConstraintsCache, a.Tracker)
 	if a.GetPod != nil {
 		r.getPod = a.GetPod
 	}
@@ -151,7 +146,6 @@ type tags struct {
 func newReconciler(
 	mgr manager.Manager,
 	cfClient *constraintclient.Client,
-	cs *watch.ControllerSwitch,
 	reporter StatsReporter,
 	constraintsCache *ConstraintsCache,
 	tracker *readiness.Tracker,
@@ -162,7 +156,6 @@ func newReconciler(
 		statusClient: mgr.GetClient(),
 		reader:       mgr.GetCache(),
 
-		cs:               cs,
 		scheme:           mgr.GetScheme(),
 		cfClient:         cfClient,
 		log:              log,
@@ -209,7 +202,6 @@ type ReconcileConstraint struct {
 	writer       client.Writer
 	statusClient client.StatusClient
 
-	cs               *watch.ControllerSwitch
 	scheme           *runtime.Scheme
 	cfClient         *constraintclient.Client
 	log              logr.Logger
@@ -229,15 +221,6 @@ type ReconcileConstraint struct {
 // Reconcile reads that state of the cluster for a constraint object and makes changes based on the state read
 // and what is in the constraint.Spec.
 func (r *ReconcileConstraint) Reconcile(ctx context.Context, request reconcile.Request) (reconcile.Result, error) {
-	// Short-circuit if shutting down.
-	if r.cs != nil {
-		running := r.cs.Enter()
-		defer r.cs.Exit()
-		if !running {
-			return reconcile.Result{}, nil
-		}
-	}
-
 	gvk, unpackedRequest, err := util.UnpackRequest(request)
 	if err != nil {
 		// Unrecoverable, do not retry.