Deploy ServiceMonitor when Promethueus operator installed

Instead of checking if we are running on an OpenShift cluster, check
if the `monitoring.coreos.com/v1` apis are available. This way,
the operator will configure prometheus depending on the
availability of the Prometheus operator.

Signed-off-by: Andrea Panattoni <apanatto@redhat.com>

bindata/manifests/metrics-exporter/metrics-service.yaml

@@ -18,7 +18,7 @@ spec:
       name: sriov-network-metrics
       port: {{ .MetricsExporterPort }}
       targetPort: {{ .MetricsExporterPort }}
-{{ if .IsOpenshift }}
+{{ if .IsPrometheusOperatorInstalled }}
 ---
 apiVersion: monitoring.coreos.com/v1
 kind: ServiceMonitor

controllers/helper.go

@@ -32,6 +32,7 @@ import (
 	"k8s.io/apimachinery/pkg/api/errors"
 	uns "k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
 	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
 	"k8s.io/apimachinery/pkg/types"
 	kscheme "k8s.io/client-go/kubernetes/scheme"
 	k8sclient "sigs.k8s.io/controller-runtime/pkg/client"
@@ -397,3 +398,21 @@ func updateDaemonsetNodeSelector(obj *uns.Unstructured, nodeSelector map[string]
 	}
 	return nil
 }
+
+func isPrometheusOperatorInstalled(ctx context.Context, client k8sclient.Reader) bool {
+	u := &uns.UnstructuredList{}
+	u.SetGroupVersionKind(schema.GroupVersionKind{
+		Group:   "monitoring.coreos.com",
+		Kind:    "ServiceMonitorList",
+		Version: "v1",
+	})
+	err := client.List(ctx, u)
+	if err != nil {
+		if errors.IsNotFound(err) {
+			return false
+		}
+		log.Log.WithName("isPrometheusOperatorInstalled").Error(err, "Error while looking for prometheus operator")
+		return false
+	}
+	return true
+}

controllers/helper_test.go

@@ -18,18 +18,24 @@ package controllers
 
 import (
 	"context"
+	"fmt"
 	"sync"
 	"testing"
 
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"
 
 	"github.com/google/go-cmp/cmp"
+	monitoringv1 "github.com/prometheus-operator/prometheus-operator/pkg/apis/monitoring/v1"
 	appsv1 "k8s.io/api/apps/v1"
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	utilruntime "k8s.io/apimachinery/pkg/util/runtime"
 	controllerruntime "sigs.k8s.io/controller-runtime"
+	k8sclient "sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
 
 	sriovnetworkv1 "github.com/k8snetworkplumbingwg/sriov-network-operator/api/v1"
 	"github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/vars"
@@ -327,4 +333,36 @@ var _ = Describe("Helper Validation", Ordered, func() {
 			Expect(in.Spec.Template.Spec.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution.NodeSelectorTerms[0].MatchExpressions[0].Key).To(Equal("test1"))
 		})
 	})
+
+	Context("isPrometheusOperatorInstalled", func() {
+		It("should return true if ServiceMonitor resource is available in the cluster", func() {
+			scheme := runtime.NewScheme()
+			utilruntime.Must(monitoringv1.AddToScheme(scheme))
+			client := fake.NewClientBuilder().
+				WithScheme(scheme).
+				Build()
+			Expect(isPrometheusOperatorInstalled(context.Background(), client)).To(BeTrue())
+		})
+
+		It("should return false if no monitoring.coreos.com resource is not available in the cluster", func() {
+			// There is no way to instruct a fake client to return an error while listing unknown resources
+			// so we use a client that returns an error on get and list
+			// https://github.com/k8snetworkplumbingwg/sriov-network-operator/blob/8d32f42b42b59a27864043ad22afa3741f197d9a/vendor/sigs.k8s.io/controller-runtime/pkg/client/fake/client.go#L513
+			client := &errorReturningClient{}
+			Expect(isPrometheusOperatorInstalled(context.Background(), client)).To(BeFalse())
+		})
+
+	})
+
 })
+
+type errorReturningClient struct {
+}
+
+func (e *errorReturningClient) Get(ctx context.Context, key k8sclient.ObjectKey, obj k8sclient.Object, opts ...k8sclient.GetOption) error {
+	return fmt.Errorf("resource get error")
+}
+
+func (e *errorReturningClient) List(ctx context.Context, list k8sclient.ObjectList, opts ...k8sclient.ListOption) error {
+	return fmt.Errorf("resource list error")
+}

controllers/sriovoperatorconfig_controller.go

@@ -232,6 +232,7 @@ func (r *SriovOperatorConfigReconciler) syncMetricsExporter(ctx context.Context,
 	data.Data["MetricsExporterPort"] = os.Getenv("METRICS_EXPORTER_PORT")
 	data.Data["MetricsExporterKubeRbacProxyImage"] = os.Getenv("METRICS_EXPORTER_KUBE_RBAC_PROXY_IMAGE")
 	data.Data["IsOpenshift"] = r.PlatformHelper.IsOpenshiftCluster()
+	data.Data["IsPrometheusOperatorInstalled"] = isPrometheusOperatorInstalled(ctx, r.Client)
 	data.Data["NodeSelectorField"] = GetDefaultNodeSelector()
 	if dc.Spec.ConfigDaemonNodeSelector != nil {
 		data.Data["NodeSelectorField"] = dc.Spec.ConfigDaemonNodeSelector

controllers/suite_test.go

@@ -41,6 +41,7 @@ import (
 	netattdefv1 "github.com/k8snetworkplumbingwg/network-attachment-definition-client/pkg/apis/k8s.cni.cncf.io/v1"
 	openshiftconfigv1 "github.com/openshift/api/config/v1"
 	mcfgv1 "github.com/openshift/machine-config-operator/pkg/apis/machineconfiguration.openshift.io/v1"
+	monitoringv1 "github.com/prometheus-operator/prometheus-operator/pkg/apis/monitoring/v1"
 
 	//+kubebuilder:scaffold:imports
 	sriovnetworkv1 "github.com/k8snetworkplumbingwg/sriov-network-operator/api/v1"
@@ -153,6 +154,8 @@ var _ = BeforeSuite(func() {
 	Expect(err).NotTo(HaveOccurred())
 	err = openshiftconfigv1.AddToScheme(scheme.Scheme)
 	Expect(err).NotTo(HaveOccurred())
+	err = monitoringv1.AddToScheme(scheme.Scheme)
+	Expect(err).NotTo(HaveOccurred())
 
 	vars.Config = cfg
 	vars.Scheme = scheme.Scheme

deploy/role.yaml

@@ -32,6 +32,7 @@ rules:
   verbs:
   - get
   - create
+  - list
 - apiGroups:
   - apps
   resourceNames:

deployment/sriov-network-operator/templates/role.yaml

@@ -35,6 +35,7 @@ rules:
     verbs:
       - get
       - create
+      - list
   - apiGroups:
       - apps
     resourceNames: