feat: Add metric reporting to ExpansionTemplate controller (#2276)

* Add metric reporting to ExpansionTemplate controller

Signed-off-by: davis-haba <davishaba@google.com>

* Update metrics docs with gatekeeper_expansion_templates metric

Signed-off-by: davis-haba <davishaba@google.com>

Signed-off-by: davis-haba <davishaba@google.com>
Co-authored-by: Rita Zhang <rita.z.zhang@gmail.com>

pkg/controller/expansion/expansion_controller.go

@@ -15,6 +15,7 @@ import (
 	"github.com/open-policy-agent/gatekeeper/pkg/watch"
 	"k8s.io/apimachinery/pkg/api/errors"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/controller"
 	"sigs.k8s.io/controller-runtime/pkg/handler"
@@ -61,12 +62,18 @@ func (a *Adder) InjectProviderCache(_ *externaldata.ProviderCache) {}
 
 type Reconciler struct {
 	client.Client
-	system *expansion.System
-	scheme *runtime.Scheme
+	system   *expansion.System
+	scheme   *runtime.Scheme
+	registry *etRegistry
 }
 
 func newReconciler(mgr manager.Manager, system *expansion.System) *Reconciler {
-	return &Reconciler{Client: mgr.GetClient(), system: system, scheme: mgr.GetScheme()}
+	return &Reconciler{
+		Client:   mgr.GetClient(),
+		system:   system,
+		scheme:   mgr.GetScheme(),
+		registry: newRegistry(),
+	}
 }
 
 func add(mgr manager.Manager, r reconcile.Reconciler) error {
@@ -97,6 +104,10 @@ func (r *Reconciler) Reconcile(ctx context.Context, request reconcile.Request) (
 	if err := r.scheme.Convert(te, unversionedTE, nil); err != nil {
 		return reconcile.Result{}, err
 	}
+	nsName := types.NamespacedName{
+		Namespace: unversionedTE.GetNamespace(),
+		Name:      unversionedTE.GetName(),
+	}
 	if deleted {
 		// unversionedTE will be an empty struct. We set the metadata name, which is
 		// used as a key to delete it from the expansion system
@@ -105,11 +116,17 @@ func (r *Reconciler) Reconcile(ctx context.Context, request reconcile.Request) (
 			return reconcile.Result{}, err
 		}
 		log.Info("removed template expansion", "template name", unversionedTE.ObjectMeta.Name)
+		r.registry.remove(nsName)
 	} else {
 		if err := r.system.UpsertTemplate(unversionedTE); err != nil {
 			return reconcile.Result{}, err
 		}
 		log.Info("upserted template expansion", "template name", unversionedTE.ObjectMeta.Name)
+		r.registry.add(nsName)
+	}
+
+	if err := r.registry.report(ctx); err != nil {
+		log.Error(err, "error reporting template expansion metrics", "namespacedName", nsName)
 	}
 
 	return reconcile.Result{}, nil

pkg/controller/expansion/stats_reporter.go

@@ -0,0 +1,70 @@
+package expansion
+
+import (
+	"context"
+
+	"github.com/open-policy-agent/gatekeeper/pkg/metrics"
+	"go.opencensus.io/stats"
+	"go.opencensus.io/stats/view"
+	"k8s.io/apimachinery/pkg/types"
+)
+
+const (
+	etMetricName = "expansion_templates"
+	etDesc       = "Number of observed expansion templates"
+)
+
+var (
+	etM = stats.Int64(etMetricName, etDesc, stats.UnitDimensionless)
+
+	views = []*view.View{
+		{
+			Name:        etMetricName,
+			Measure:     etM,
+			Description: etDesc,
+			Aggregation: view.LastValue(),
+		},
+	}
+)
+
+func init() {
+	if err := register(); err != nil {
+		panic(err)
+	}
+}
+
+func register() error {
+	return view.Register(views...)
+}
+
+func newRegistry() *etRegistry {
+	return &etRegistry{cache: make(map[types.NamespacedName]bool)}
+}
+
+type etRegistry struct {
+	cache map[types.NamespacedName]bool
+	dirty bool
+}
+
+func (r *etRegistry) add(key types.NamespacedName) {
+	r.cache[key] = true
+	r.dirty = true
+}
+
+func (r *etRegistry) remove(key types.NamespacedName) {
+	delete(r.cache, key)
+	r.dirty = true
+}
+
+func (r *etRegistry) report(ctx context.Context) error {
+	if !r.dirty {
+		return nil
+	}
+
+	if err := metrics.Record(ctx, etM.M(int64(len(r.cache)))); err != nil {
+		r.dirty = false
+		return err
+	}
+
+	return nil
+}

website/docs/metrics.md

@@ -51,6 +51,14 @@ Below are the list of metrics provided by Gatekeeper:
 
     Aggregation: `Distribution`
 
+## Expansion Template
+
+- Name: `gatekeeper_expansion_templates`
+
+  Description: `Number of observed expansion templates`
+
+  Aggregation: `LastValue`
+
 ## Webhook
 
 - Name: `gatekeeper_validation_request_count`