feat(go): track gateway active connections

docs/ops/go-service-observability.md

@@ -31,6 +31,7 @@ curl -fsS http://127.0.0.1:20200/readyz
 
 | Metric | Meaning |
 | --- | --- |
+| `vehicle_gateway_active_connections` | Current TCP connections by protocol. Labels: `protocol`. |
 | `vehicle_gateway_frames_total` | Protocol frames received and parsed by the gateway. Labels: `protocol`, `status`. |
 | `vehicle_gateway_publish_total` | Gateway publish result for raw and unified events. Labels: `protocol`, `kind`, `status`. |
 | `vehicle_bridge_messages_total` | NATS messages fetched by the bridge. Labels: `subject`, `status`. |

go/vehicle-gateway/internal/gateway/tcp_server.go

@@ -145,6 +145,8 @@ func (s *TCPServer) handleConnection(ctx context.Context, conn net.Conn) {
 	defer conn.Close()
 	source := conn.RemoteAddr().String()
 	log := s.logger.With("protocol", s.protocol.Protocol, "remote", source)
+	s.recordConnectionMetric(1)
+	defer s.recordConnectionMetric(-1)
 	log.Info("tcp connection opened")
 	defer log.Info("tcp connection closed")
 
@@ -268,6 +270,15 @@ func (s *TCPServer) recordPublishMetric(kind string, status string) {
 	})
 }
 
+func (s *TCPServer) recordConnectionMetric(delta float64) {
+	if s.metrics == nil {
+		return
+	}
+	s.metrics.AddGauge("vehicle_gateway_active_connections", metrics.Labels{
+		"protocol": string(s.protocol.Protocol),
+	}, delta)
+}
+
 func (p TCPProtocol) String() string {
 	return fmt.Sprintf("%s@%s", p.Protocol, p.Addr)
 }

go/vehicle-gateway/internal/gateway/tcp_server_test.go

@@ -88,6 +88,51 @@ func TestTCPServerRecordsFrameMetrics(t *testing.T) {
 	}
 }
 
+func TestTCPServerRecordsActiveConnectionGauge(t *testing.T) {
+	registry := metrics.NewRegistry()
+	server, err := NewTCPServer(TCPServerConfig{
+		Protocol: TCPProtocol{
+			Protocol: envelope.ProtocolJT808,
+			Addr:     ":0",
+			Extract:  jt808.ExtractFrames,
+			Parse:    jt808.ParseFrame,
+		},
+		Sink:        &recordingSink{},
+		Logger:      slog.New(slog.NewTextHandler(testWriter{t: t}, nil)),
+		Metrics:     registry,
+		IdleTimeout: time.Minute,
+	})
+	if err != nil {
+		t.Fatalf("NewTCPServer() error = %v", err)
+	}
+	client, srv := net.Pipe()
+	opened := make(chan struct{})
+	done := make(chan struct{})
+	go func() {
+		close(opened)
+		server.handleConnection(context.Background(), srv)
+		close(done)
+	}()
+	<-opened
+	deadline := time.Now().Add(2 * time.Second)
+	for time.Now().Before(deadline) {
+		if strings.Contains(registry.Render(), `vehicle_gateway_active_connections{protocol="JT808"} 1`) {
+			break
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+	if text := registry.Render(); !strings.Contains(text, `vehicle_gateway_active_connections{protocol="JT808"} 1`) {
+		t.Fatalf("active connection gauge missing while open:\n%s", text)
+	}
+
+	_ = client.Close()
+	<-done
+
+	if text := registry.Render(); !strings.Contains(text, `vehicle_gateway_active_connections{protocol="JT808"} 0`) {
+		t.Fatalf("active connection gauge should return to zero:\n%s", text)
+	}
+}
+
 func TestTCPServerPublishesBadFrameOnlyToRaw(t *testing.T) {
 	good := buildGBFrame(0x02, 0xfe, "LNBSCB3D4R1234567", nil)
 	good[len(good)-1] ^= 0xff

go/vehicle-gateway/internal/metrics/metrics.go

@@ -13,10 +13,14 @@ type Labels map[string]string
 type Registry struct {
 	mu       sync.RWMutex
 	counters map[string]map[string]float64
+	gauges   map[string]map[string]float64
 }
 
 func NewRegistry() *Registry {
-	return &Registry{counters: map[string]map[string]float64{}}
+	return &Registry{
+		counters: map[string]map[string]float64{},
+		gauges:   map[string]map[string]float64{},
+	}
 }
 
 func (r *Registry) IncCounter(name string, labels Labels) {
@@ -37,6 +41,34 @@ func (r *Registry) AddCounter(name string, labels Labels, value float64) {
 	r.counters[name][key] += value
 }
 
+func (r *Registry) SetGauge(name string, labels Labels, value float64) {
+	name = strings.TrimSpace(name)
+	if r == nil || name == "" {
+		return
+	}
+	key := labelsKey(labels)
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	if r.gauges[name] == nil {
+		r.gauges[name] = map[string]float64{}
+	}
+	r.gauges[name][key] = value
+}
+
+func (r *Registry) AddGauge(name string, labels Labels, value float64) {
+	name = strings.TrimSpace(name)
+	if r == nil || name == "" || value == 0 {
+		return
+	}
+	key := labelsKey(labels)
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	if r.gauges[name] == nil {
+		r.gauges[name] = map[string]float64{}
+	}
+	r.gauges[name][key] += value
+}
+
 func (r *Registry) Render() string {
 	if r == nil {
 		return ""
@@ -52,11 +84,28 @@ func (r *Registry) Render() string {
 
 	var b strings.Builder
 	for _, name := range names {
+		renderMetricFamily(&b, name, "counter", r.counters[name])
+	}
+
+	names = names[:0]
+	for name := range r.gauges {
+		names = append(names, name)
+	}
+	sort.Strings(names)
+	for _, name := range names {
+		renderMetricFamily(&b, name, "gauge", r.gauges[name])
+	}
+	return b.String()
+}
+
+func renderMetricFamily(b *strings.Builder, name string, metricType string, values map[string]float64) {
 	b.WriteString("# TYPE ")
 	b.WriteString(name)
-		b.WriteString(" counter\n")
+	b.WriteString(" ")
+	b.WriteString(metricType)
+	b.WriteString("\n")
 	var keys []string
-		for key := range r.counters[name] {
+	for key := range values {
 		keys = append(keys, key)
 	}
 	sort.Strings(keys)
@@ -68,11 +117,9 @@ func (r *Registry) Render() string {
 			b.WriteString("}")
 		}
 		b.WriteString(" ")
-			b.WriteString(formatNumber(r.counters[name][key]))
+		b.WriteString(formatNumber(values[key]))
 		b.WriteString("\n")
 	}
-	}
-	return b.String()
 }
 
 func NewHandler(registry *Registry) http.Handler {

go/vehicle-gateway/internal/metrics/metrics_test.go

@@ -27,6 +27,25 @@ func TestRegistryRendersCountersWithLabels(t *testing.T) {
 	}
 }
 
+func TestRegistryRendersGaugesWithLabels(t *testing.T) {
+	registry := NewRegistry()
+	registry.SetGauge("vehicle_gateway_connections", Labels{"protocol": "JT808"}, 2)
+	registry.AddGauge("vehicle_gateway_connections", Labels{"protocol": "JT808"}, -1)
+	registry.SetGauge("vehicle_gateway_connections", Labels{"protocol": "GB32960"}, 3)
+
+	text := registry.Render()
+
+	for _, want := range []string{
+		`# TYPE vehicle_gateway_connections gauge`,
+		`vehicle_gateway_connections{protocol="GB32960"} 3`,
+		`vehicle_gateway_connections{protocol="JT808"} 1`,
+	} {
+		if !strings.Contains(text, want) {
+			t.Fatalf("metrics missing %s:\n%s", want, text)
+		}
+	}
+}
+
 func TestHandlerServesPrometheusText(t *testing.T) {
 	registry := NewRegistry()
 	registry.IncCounter("vehicle_kafka_commits_total", Labels{"service": "history"})