coordinator: make metrics endpoint configurable

coordinator/main.go

@@ -19,6 +19,10 @@ import (
 	"golang.org/x/sync/errgroup"
 )
 
+const (
+	metricsPortEnvVar = "CONTRAST_METRICS_PORT"
+)
+
 func main() {
 	if err := run(); err != nil {
 		os.Exit(1)
@@ -43,6 +47,8 @@ func run() (retErr error) {
 		return fmt.Errorf("setting up mount: %w", err)
 	}
 
+	metricsPort := os.Getenv(metricsPortEnvVar)
+
 	caInstance, err := ca.New()
 	if err != nil {
 		return fmt.Errorf("creating CA: %w", err)
@@ -57,15 +63,21 @@ func run() (retErr error) {
 	eg := errgroup.Group{}
 
 	eg.Go(func() error {
-		logger.Info("Starting prometheus /metrics endpoint")
+		if metricsPort == "" {
+			return nil
+		}
+		if metricsPort == userapi.Port || metricsPort == meshapi.Port {
+			return fmt.Errorf("invalid port for metrics endpoint: %s", metricsPort)
+		}
+		logger.Info("Starting prometheus /metrics endpoint on port " + metricsPort)
 		mux := http.NewServeMux()
 		mux.Handle("/metrics", promhttp.InstrumentMetricHandler(
 			promRegistry, promhttp.HandlerFor(
 				promRegistry,
 				promhttp.HandlerOpts{Registry: promRegistry},
 			),
 		))
-		if err := http.ListenAndServe(":9102", mux); err != nil {
+		if err := http.ListenAndServe(":"+metricsPort, mux); err != nil {
 			return fmt.Errorf("serving Prometheus endpoint: %w", err)
 		}
 		return nil

docs/docs/architecture/observability.md

@@ -1,18 +1,17 @@
 # Observability
 
-The Contrast Coordinator exposes metrics in the
+The Contrast Coordinator can expose metrics in the
 [Prometheus](https://prometheus.io/) format. These can be monitored to quickly
 identify problems in the gRPC layer or attestation errors. Prometheus metrics
 are numerical values associated with a name and additional key/values pairs,
 called labels.
 
 ## Exposed metrics
 
-The Coordinator pod has the annotation `prometheus.io/scrape` set to `true` so
-it can be found by the [service discovery of
-Prometheus](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#kubernetes_sd_config).
-The metrics can be accessed at the Coordinator pod at port `9102` under the
-`/metrics` endpoint.
+The metrics can be accessed at the Coordinator pod at the port specified in the
+`CONTRAST_METRICS_PORT` environment variable under the `/metrics` endpoint. By
+default, this environment variable isn't specified, hence no metrics will be
+exposed.
 
 The Coordinator starts two gRPC servers, one for the user API on port `1313` and
 one for the mesh API on port `7777`. Metrics for both servers can be accessed

e2e/internal/contrasttest/contrasttest.go

@@ -96,6 +96,7 @@ func (ct *ContrastTest) Init(t *testing.T, resources []any) {
 	resources = kuberesource.PatchImages(resources, ct.ImageReplacements)
 	resources = kuberesource.PatchNamespaces(resources, ct.Namespace)
 	resources = kuberesource.PatchServiceMeshAdminInterface(resources, 9901)
+	resources = kuberesource.PatchCoordinatorMetrics(resources, 9102)
 	resources = kuberesource.AddLogging(resources, "debug")
 	unstructuredResources, err := kuberesource.ResourcesToUnstructured(resources)
 	require.NoError(err)

e2e/openssl/openssl_test.go

@@ -12,6 +12,7 @@ import (
 	"encoding/json"
 	"flag"
 	"log"
+	"net"
 	"os"
 	"testing"
 	"time"
@@ -64,7 +65,12 @@ func TestOpenSSL(t *testing.T) {
 		require.NoError(err)
 		require.Len(frontendPods, 1, "pod not found: %s/%s", ct.Namespace, opensslFrontend)
 
-		_, stderr, err := ct.Kubeclient.Exec(ctx, ct.Namespace, frontendPods[0].Name, []string{"/bin/bash", "-c", "curl --fail coordinator:9102/metrics"})
+		coordinatorPods, err := ct.Kubeclient.PodsFromOwner(ctx, ct.Namespace, "StatefulSet", "coordinator")
+		require.NoError(err)
+		require.NotEmpty(coordinatorPods, "pod not found: %s/%s", ct.Namespace, "coordinator")
+
+		argv := []string{"/bin/bash", "-c", "curl --fail " + net.JoinHostPort(coordinatorPods[0].Status.PodIP, "9102") + "/metrics"}
+		_, stderr, err := ct.Kubeclient.Exec(ctx, ct.Namespace, frontendPods[0].Name, argv)
 		require.NoError(err, "stderr: %q", stderr)
 	})
 

internal/kuberesource/mutators.go

@@ -12,7 +12,10 @@ import (
 	applycorev1 "k8s.io/client-go/applyconfigurations/core/v1"
 )
 
-const exposeServiceAnnotation = "contrast.edgeless.systems/expose-service"
+const (
+	exposeServiceAnnotation   = "contrast.edgeless.systems/expose-service"
+	contrastRoleAnnotationKey = "contrast.edgeless.systems/pod-role"
+)
 
 // AddInitializer adds an initializer and its shared volume to the resource.
 //
@@ -194,6 +197,24 @@ func PatchServiceMeshAdminInterface(resources []any, port int32) []any {
 	return resources
 }
 
+// PatchCoordinatorMetrics enables Coordinator metrics on the specified port.
+func PatchCoordinatorMetrics(resources []any, port int32) []any {
+	for _, resource := range resources {
+		switch r := resource.(type) {
+		case *applyappsv1.StatefulSetApplyConfiguration:
+			if r.Spec.Template.Annotations[contrastRoleAnnotationKey] == "coordinator" {
+				r.Spec.Template.Spec.Containers[0].WithEnv(NewEnvVar("CONTRAST_METRICS_PORT", fmt.Sprint(port)))
+				r.Spec.Template.Spec.Containers[0].WithPorts(
+					ContainerPort().
+						WithName("prometheus").
+						WithContainerPort(port),
+				)
+			}
+		}
+	}
+	return resources
+}
+
 // MapPodSpec applies a function to a PodSpec in a Kubernetes resource.
 func MapPodSpec(resource any, f func(spec *applycorev1.PodSpecApplyConfiguration) *applycorev1.PodSpecApplyConfiguration) any {
 	if resource == nil {

internal/kuberesource/parts.go

@@ -147,7 +147,7 @@ func Coordinator(namespace string) *CoordinatorConfig {
 				WithWhenScaled(appsv1.DeletePersistentVolumeClaimRetentionPolicyType)). // TODO(burgerdev): this should be RETAIN for released coordinators.
 			WithTemplate(PodTemplateSpec().
 				WithLabels(map[string]string{"app.kubernetes.io/name": "coordinator"}).
-				WithAnnotations(map[string]string{"contrast.edgeless.systems/pod-role": "coordinator", "prometheus.io/scrape": "true"}).
+				WithAnnotations(map[string]string{"contrast.edgeless.systems/pod-role": "coordinator"}).
 				WithSpec(PodSpec().
 					WithRuntimeClassName(runtimeHandler).
 					WithContainers(
@@ -170,9 +170,6 @@ func Coordinator(namespace string) *CoordinatorConfig {
 								ContainerPort().
 									WithName("meshapi").
 									WithContainerPort(7777),
-								ContainerPort().
-									WithName("prometheus").
-									WithContainerPort(9102),
 							).
 							WithReadinessProbe(Probe().
 								WithInitialDelaySeconds(1).