[client] Add local Prometheus metrics endpoint (#6689)

client/internal/metrics/influxdb.go

@@ -45,30 +45,13 @@ func (m *influxDBMetrics) RecordConnectionStages(
 	isReconnection bool,
 	timestamps ConnectionStageTimestamps,
 ) {
-	var signalingReceivedToConnection, connectionToWgHandshake, totalDuration float64
-
-	if !timestamps.SignalingReceived.IsZero() && !timestamps.ConnectionReady.IsZero() {
-		signalingReceivedToConnection = timestamps.ConnectionReady.Sub(timestamps.SignalingReceived).Seconds()
-	}
-
-	if !timestamps.ConnectionReady.IsZero() && !timestamps.WgHandshakeSuccess.IsZero() {
-		connectionToWgHandshake = timestamps.WgHandshakeSuccess.Sub(timestamps.ConnectionReady).Seconds()
-	}
-
-	if !timestamps.SignalingReceived.IsZero() && !timestamps.WgHandshakeSuccess.IsZero() {
-		totalDuration = timestamps.WgHandshakeSuccess.Sub(timestamps.SignalingReceived).Seconds()
-	}
-
-	attemptType := "initial"
-	if isReconnection {
-		attemptType = "reconnection"
-	}
+	signalingReceivedToConnection, connectionToWgHandshake, totalDuration := timestamps.Durations()
 
 	connTypeStr := connectionType.String()
 	tags := fmt.Sprintf("deployment_type=%s,connection_type=%s,attempt_type=%s,version=%s,os=%s,arch=%s,peer_id=%s,connection_pair_id=%s",
 		agentInfo.DeploymentType.String(),
 		connTypeStr,
-		attemptType,
+		attemptType(isReconnection),
 		agentInfo.Version,
 		agentInfo.OS,
 		agentInfo.Arch,
@@ -94,7 +77,7 @@ func (m *influxDBMetrics) RecordConnectionStages(
 	m.trimLocked()
 
 	log.Tracef("peer connection metrics [%s, %s, %s]: signalingReceived→connection: %.3fs, connection→wg_handshake: %.3fs, total: %.3fs",
-		agentInfo.DeploymentType.String(), connTypeStr, attemptType, signalingReceivedToConnection, connectionToWgHandshake, totalDuration)
+		agentInfo.DeploymentType.String(), connTypeStr, attemptType(isReconnection), signalingReceivedToConnection, connectionToWgHandshake, totalDuration)
 }
 
 func (m *influxDBMetrics) RecordSyncDuration(_ context.Context, agentInfo AgentInfo, duration time.Duration) {

client/internal/metrics/metrics.go

@@ -89,6 +89,21 @@ type ConnectionStageTimestamps struct {
 	WgHandshakeSuccess time.Time
 }
 
+// Durations returns the stage durations in seconds. A duration is zero when
+// either of its timestamps is missing.
+func (c ConnectionStageTimestamps) Durations() (signalingToConnection, connectionToWgHandshake, total float64) {
+	if !c.SignalingReceived.IsZero() && !c.ConnectionReady.IsZero() {
+		signalingToConnection = c.ConnectionReady.Sub(c.SignalingReceived).Seconds()
+	}
+	if !c.ConnectionReady.IsZero() && !c.WgHandshakeSuccess.IsZero() {
+		connectionToWgHandshake = c.WgHandshakeSuccess.Sub(c.ConnectionReady).Seconds()
+	}
+	if !c.SignalingReceived.IsZero() && !c.WgHandshakeSuccess.IsZero() {
+		total = c.WgHandshakeSuccess.Sub(c.SignalingReceived).Seconds()
+	}
+	return signalingToConnection, connectionToWgHandshake, total
+}
+
 // String returns a human-readable representation of the connection stage timestamps
 func (c ConnectionStageTimestamps) String() string {
 	return fmt.Sprintf("ConnectionStageTimestamps{SignalingReceived=%v, ConnectionReady=%v, WgHandshakeSuccess=%v}",
@@ -279,3 +294,11 @@ func (c *ClientMetrics) stopPushLocked() {
 	c.wg.Wait()
 	c.push.Store(nil)
 }
+
+// attemptType returns the metric label for an initial vs reconnection attempt.
+func attemptType(isReconnection bool) string {
+	if isReconnection {
+		return "reconnection"
+	}
+	return "initial"
+}

client/internal/metrics/metrics_default.go

@@ -2,10 +2,24 @@
 
 package metrics
 
+import "github.com/prometheus/client_golang/prometheus"
+
 // NewClientMetrics creates a new ClientMetrics instance
 func NewClientMetrics(agentInfo AgentInfo) *ClientMetrics {
 	return &ClientMetrics{
-		impl:      newInfluxDBMetrics(),
+		impl:      newPrometheusMetrics(newInfluxDBMetrics()),
 		agentInfo: agentInfo,
 	}
 }
+
+// PrometheusGatherer returns the registry with the mirrored Prometheus
+// metrics, or nil when unavailable.
+func (c *ClientMetrics) PrometheusGatherer() prometheus.Gatherer {
+	if c == nil {
+		return nil
+	}
+	if pm, ok := c.impl.(*prometheusMetrics); ok {
+		return pm.Gatherer()
+	}
+	return nil
+}

client/internal/metrics/prometheus.go

@@ -0,0 +1,119 @@
+//go:build !js
+
+package metrics
+
+import (
+	"context"
+	"io"
+	"strconv"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+// prometheusMetrics mirrors recorded client metrics into a Prometheus
+// registry for the local /metrics endpoint, then delegates to the wrapped
+// implementation. Export and Reset pass through untouched: Prometheus
+// metrics are cumulative and pull-based.
+type prometheusMetrics struct {
+	next     metricsImplementation
+	registry *prometheus.Registry
+
+	connectionStages  *prometheus.HistogramVec
+	syncDuration      prometheus.Histogram
+	syncPhaseDuration *prometheus.HistogramVec
+	loginDuration     *prometheus.HistogramVec
+}
+
+func newPrometheusMetrics(next metricsImplementation) *prometheusMetrics {
+	connectionBuckets := []float64{.05, .1, .25, .5, 1, 2.5, 5, 10, 30, 60}
+
+	m := &prometheusMetrics{
+		next:     next,
+		registry: prometheus.NewRegistry(),
+		connectionStages: prometheus.NewHistogramVec(prometheus.HistogramOpts{
+			Name:    "netbird_peer_connection_stage_duration_seconds",
+			Help:    "Duration of peer connection establishment stages.",
+			Buckets: connectionBuckets,
+		}, []string{"stage", "connection_type", "attempt_type"}),
+		syncDuration: prometheus.NewHistogram(prometheus.HistogramOpts{
+			Name:    "netbird_sync_duration_seconds",
+			Help:    "Duration of management sync message processing.",
+			Buckets: prometheus.DefBuckets,
+		}),
+		syncPhaseDuration: prometheus.NewHistogramVec(prometheus.HistogramOpts{
+			Name:    "netbird_sync_phase_duration_seconds",
+			Help:    "Duration of individual sync processing phases.",
+			Buckets: prometheus.DefBuckets,
+		}, []string{"phase"}),
+		loginDuration: prometheus.NewHistogramVec(prometheus.HistogramOpts{
+			Name:    "netbird_login_duration_seconds",
+			Help:    "Duration of logins to the management service.",
+			Buckets: prometheus.DefBuckets,
+		}, []string{"success"}),
+	}
+
+	m.registry.MustRegister(m.connectionStages, m.syncDuration, m.syncPhaseDuration, m.loginDuration)
+	return m
+}
+
+// Gatherer returns the registry holding the mirrored metrics.
+func (m *prometheusMetrics) Gatherer() prometheus.Gatherer {
+	return m.registry
+}
+
+// RecordConnectionStages implements metricsImplementation.
+func (m *prometheusMetrics) RecordConnectionStages(
+	ctx context.Context,
+	agentInfo AgentInfo,
+	connectionPairID string,
+	connectionType ConnectionType,
+	isReconnection bool,
+	timestamps ConnectionStageTimestamps,
+) {
+	attempt := attemptType(isReconnection)
+	connType := connectionType.String()
+
+	signalingToConnection, connectionToWgHandshake, total := timestamps.Durations()
+	if signalingToConnection > 0 {
+		m.connectionStages.WithLabelValues("signaling_to_connection", connType, attempt).Observe(signalingToConnection)
+	}
+	if connectionToWgHandshake > 0 {
+		m.connectionStages.WithLabelValues("connection_to_wg_handshake", connType, attempt).Observe(connectionToWgHandshake)
+	}
+	if total > 0 {
+		m.connectionStages.WithLabelValues("total", connType, attempt).Observe(total)
+	}
+
+	m.next.RecordConnectionStages(ctx, agentInfo, connectionPairID, connectionType, isReconnection, timestamps)
+}
+
+// RecordSyncDuration implements metricsImplementation.
+func (m *prometheusMetrics) RecordSyncDuration(ctx context.Context, agentInfo AgentInfo, duration time.Duration) {
+	m.syncDuration.Observe(duration.Seconds())
+	m.next.RecordSyncDuration(ctx, agentInfo, duration)
+}
+
+// RecordSyncPhase implements metricsImplementation.
+func (m *prometheusMetrics) RecordSyncPhase(ctx context.Context, agentInfo AgentInfo, phase string, duration time.Duration) {
+	m.syncPhaseDuration.WithLabelValues(phase).Observe(duration.Seconds())
+	m.next.RecordSyncPhase(ctx, agentInfo, phase, duration)
+}
+
+// RecordLoginDuration implements metricsImplementation.
+func (m *prometheusMetrics) RecordLoginDuration(ctx context.Context, agentInfo AgentInfo, duration time.Duration, success bool) {
+	m.loginDuration.WithLabelValues(strconv.FormatBool(success)).Observe(duration.Seconds())
+	m.next.RecordLoginDuration(ctx, agentInfo, duration, success)
+}
+
+// Export implements metricsImplementation by delegating to the wrapped
+// implementation; Prometheus metrics are pulled via the registry instead.
+func (m *prometheusMetrics) Export(w io.Writer) error {
+	return m.next.Export(w)
+}
+
+// Reset implements metricsImplementation by delegating to the wrapped
+// implementation; Prometheus metrics must not be cleared on push.
+func (m *prometheusMetrics) Reset() {
+	m.next.Reset()
+}