feat(metrics): provide metrics for tenant quotas (#1094)

Signed-off-by: Lukas Boettcher <1340215+lukasboettcher@users.noreply.github.com>

controllers/tenant/manager.go

@@ -20,6 +20,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/reconcile"
 
 	capsulev1beta2 "github.com/projectcapsule/capsule/api/v1beta2"
+	"github.com/projectcapsule/capsule/pkg/metrics"
 )
 
 type Manager struct {
@@ -49,6 +50,10 @@ func (r Manager) Reconcile(ctx context.Context, request ctrl.Request) (result ct
 		if apierrors.IsNotFound(err) {
 			r.Log.Info("Request object not found, could have been deleted after reconcile request")
 
+			// If tenant was deleted or cannot be found, clean up metrics
+			metrics.TenantResourceUsage.DeletePartialMatch(map[string]string{"tenant": request.Name})
+			metrics.TenantResourceLimit.DeletePartialMatch(map[string]string{"tenant": request.Name})
+
 			return reconcile.Result{}, nil
 		}
 

controllers/tenant/resourcequotas.go

@@ -23,6 +23,7 @@ import (
 
 	capsulev1beta2 "github.com/projectcapsule/capsule/api/v1beta2"
 	"github.com/projectcapsule/capsule/pkg/api"
+	"github.com/projectcapsule/capsule/pkg/metrics"
 	"github.com/projectcapsule/capsule/pkg/utils"
 )
 
@@ -51,6 +52,18 @@ func (r *Manager) syncResourceQuotas(ctx context.Context, tenant *capsulev1beta2
 	if typeLabel, err = utils.GetTypeLabel(&corev1.ResourceQuota{}); err != nil {
 		return err
 	}
+
+	// Remove prior metrics, to avoid cleaning up for metrics of deleted ResourceQuotas
+	metrics.TenantResourceUsage.DeletePartialMatch(map[string]string{"tenant": tenant.Name})
+	metrics.TenantResourceLimit.DeletePartialMatch(map[string]string{"tenant": tenant.Name})
+
+	// Expose the namespace quota and usage as metrics for the tenant
+	metrics.TenantResourceUsage.WithLabelValues(tenant.Name, "namespaces", "").Set(float64(tenant.Status.Size))
+
+	if tenant.Spec.NamespaceOptions != nil && tenant.Spec.NamespaceOptions.Quota != nil {
+		metrics.TenantResourceLimit.WithLabelValues(tenant.Name, "namespaces", "").Set(float64(*tenant.Spec.NamespaceOptions.Quota))
+	}
+
 	//nolint:nestif
 	if tenant.Spec.ResourceQuota.Scope == api.ResourceQuotaScopeTenant {
 		group := new(errgroup.Group)
@@ -102,6 +115,19 @@ func (r *Manager) syncResourceQuotas(ctx context.Context, tenant *capsulev1beta2
 
 					r.Log.Info("Computed " + name.String() + " quota for the whole Tenant is " + quantity.String())
 
+					// Expose usage and limit metrics for the resource (name) of the ResourceQuota (index)
+					metrics.TenantResourceUsage.WithLabelValues(
+						tenant.Name,
+						name.String(),
+						strconv.Itoa(index),
+					).Set(float64(quantity.MilliValue()) / 1000)
+
+					metrics.TenantResourceLimit.WithLabelValues(
+						tenant.Name,
+						name.String(),
+						strconv.Itoa(index),
+					).Set(float64(hardQuota.MilliValue()) / 1000)
+
 					switch quantity.Cmp(resourceQuota.Hard[name]) {
 					case 0:
 						// The Tenant is matching exactly the Quota:

pkg/metrics/metrics.go

@@ -0,0 +1,30 @@
+// Copyright 2020-2023 Project Capsule Authors.
+// SPDX-License-Identifier: Apache-2.0
+
+package metrics
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"sigs.k8s.io/controller-runtime/pkg/metrics"
+)
+
+var (
+	metricsPrefix = "capsule_"
+
+	TenantResourceUsage = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Name: metricsPrefix + "tenant_resource_usage",
+		Help: "Current resource usage for a given resource in a tenant",
+	}, []string{"tenant", "resource", "resourcequotaindex"})
+
+	TenantResourceLimit = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Name: metricsPrefix + "tenant_resource_limit",
+		Help: "Current resource limit for a given resource in a tenant",
+	}, []string{"tenant", "resource", "resourcequotaindex"})
+)
+
+func init() {
+	metrics.Registry.MustRegister(
+		TenantResourceUsage,
+		TenantResourceLimit,
+	)
+}