Fix missing metrics and verify metrics are not lost

Signed-off-by: Davanum Srinivas <davanum@gmail.com>

server/config/config.go

@@ -211,6 +211,9 @@ type ServerConfig struct {
 
 	// ServerFeatureGate is a server level feature gate
 	ServerFeatureGate featuregate.FeatureGate
+
+	// Metrics types of metrics - should be either 'basic' or 'extensive'
+	Metrics string
 }
 
 // VerifyBootstrap sanity-checks the initial config for bootstrap case

server/embed/etcd.go

@@ -31,7 +31,6 @@ import (
 	"sync"
 	"time"
 
-	"github.com/prometheus/client_golang/prometheus"
 	"github.com/soheilhy/cmux"
 	"go.uber.org/zap"
 	"google.golang.org/grpc"
@@ -227,6 +226,7 @@ func StartEtcd(inCfg *Config) (e *Etcd, err error) {
 		V2Deprecation:                                 cfg.V2DeprecationEffective(),
 		ExperimentalLocalAddress:                      cfg.InferLocalAddr(),
 		ServerFeatureGate:                             cfg.ServerFeatureGate,
+		Metrics:                                       cfg.Metrics,
 	}
 
 	if srvcfg.ExperimentalEnableDistributedTracing {
@@ -844,18 +844,6 @@ func (e *Etcd) createMetricsListener(murl url.URL) (net.Listener, error) {
 }
 
 func (e *Etcd) serveMetrics() (err error) {
-	if e.cfg.Metrics == "extensive" {
-		var opts prometheus.HistogramOpts
-		serverHandledHistogram := prometheus.NewHistogramVec(
-			opts,
-			[]string{"grpc_type", "grpc_service", "grpc_method"},
-		)
-		err := prometheus.Register(serverHandledHistogram)
-		if err != nil {
-			e.GetLogger().Error("setting up prometheus metrics failed.", zap.Error(err))
-		}
-	}
-
 	if len(e.cfg.ListenMetricsUrls) > 0 {
 		metricsMux := http.NewServeMux()
 		etcdhttp.HandleMetrics(metricsMux)

server/etcdserver/api/v3rpc/grpc.go

@@ -58,6 +58,13 @@ func Server(s *etcdserver.EtcdServer, tls *tls.Config, interceptor grpc.UnarySer
 		serverMetrics.StreamServerInterceptor(),
 	}
 
+	// If extensive metrics are enabled, register a histogram to track the reponse latency of gRPC requests
+	if s.Cfg.Metrics == "extensive" {
+		unaryInterceptor, streamInterceptor := constructExtensiveMetricsInterceptors()
+		chainUnaryInterceptors = append(chainUnaryInterceptors, unaryInterceptor)
+		chainStreamInterceptors = append(chainStreamInterceptors, streamInterceptor)
+	}
+
 	if s.Cfg.ExperimentalEnableDistributedTracing {
 		chainUnaryInterceptors = append(chainUnaryInterceptors, otelgrpc.UnaryServerInterceptor(s.Cfg.ExperimentalTracerOptions...))
 		chainStreamInterceptors = append(chainStreamInterceptors, otelgrpc.StreamServerInterceptor(s.Cfg.ExperimentalTracerOptions...))

server/etcdserver/api/v3rpc/metrics.go

@@ -14,7 +14,14 @@
 
 package v3rpc
 
-import "github.com/prometheus/client_golang/prometheus"
+import (
+	"context"
+	"strings"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"google.golang.org/grpc"
+)
 
 var (
 	sentBytes = prometheus.NewCounter(prometheus.CounterOpts{
@@ -56,3 +63,46 @@ func init() {
 	prometheus.MustRegister(streamFailures)
 	prometheus.MustRegister(clientRequests)
 }
+
+func splitMethodName(fullMethodName string) (string, string) {
+	fullMethodName = strings.TrimPrefix(fullMethodName, "/") // remove leading slash
+	if i := strings.Index(fullMethodName, "/"); i >= 0 {
+		return fullMethodName[:i], fullMethodName[i+1:]
+	}
+	return "unknown", "unknown"
+}
+
+// constructExtensiveMetricsInterceptors constructs unary and stream interceptors to record histogram metrics for gRPC requests
+func constructExtensiveMetricsInterceptors() (grpc.UnaryServerInterceptor, grpc.StreamServerInterceptor) {
+	// Define a new histogram metric using default buckets
+	serverHandledHistogram := prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Name:    "grpc_server_handling_seconds",
+			Help:    "Histogram of response latency (seconds) of gRPC that had been application-level handled by the server.",
+			Buckets: prometheus.DefBuckets,
+		},
+		[]string{"grpc_type", "grpc_service", "grpc_method"},
+	)
+	prometheus.Register(serverHandledHistogram)
+
+	// method to record histogram metrics for both unary and stream requests
+	recordHistogramMetrics := func(serverHandledHistogram *prometheus.HistogramVec, grpcType, fullMethodName string, startTime time.Time) {
+		grpcService, grpcMethod := splitMethodName(fullMethodName)
+		serverHandledHistogram.WithLabelValues(grpcType, grpcService, grpcMethod).Observe(time.Since(startTime).Seconds())
+	}
+
+	// Add a new interceptor to spit out histogram metrics for unary requests
+	unaryInterceptor := func(ctx context.Context, req any, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (resp any, err error) {
+		startTime := time.Now()
+		resp, err = handler(ctx, req)
+		recordHistogramMetrics(serverHandledHistogram, "unary", info.FullMethod, startTime)
+		return resp, err
+	}
+	streamInterceptor := func(srv any, ss grpc.ServerStream, info *grpc.StreamServerInfo, handler grpc.StreamHandler) error {
+		startTime := time.Now()
+		err := handler(srv, ss)
+		recordHistogramMetrics(serverHandledHistogram, "stream", info.FullMethod, startTime)
+		return err
+	}
+	return unaryInterceptor, streamInterceptor
+}

tests/framework/integration/cluster.go

@@ -175,6 +175,7 @@ type ClusterConfig struct {
 	ExperimentalMaxLearners     int
 	DisableStrictReconfigCheck  bool
 	CorruptCheckTime            time.Duration
+	Metrics                     string
 }
 
 type Cluster struct {
@@ -292,6 +293,7 @@ func (c *Cluster) MustNewMember(t testutil.TB) *Member {
 			ExperimentalMaxLearners:     c.Cfg.ExperimentalMaxLearners,
 			DisableStrictReconfigCheck:  c.Cfg.DisableStrictReconfigCheck,
 			CorruptCheckTime:            c.Cfg.CorruptCheckTime,
+			Metrics:                     c.Cfg.Metrics,
 		})
 	m.DiscoveryURL = c.Cfg.DiscoveryURL
 	return m
@@ -617,6 +619,7 @@ type MemberConfig struct {
 	ExperimentalMaxLearners     int
 	DisableStrictReconfigCheck  bool
 	CorruptCheckTime            time.Duration
+	Metrics                     string
 }
 
 // MustNewMember return an inited member with the given name. If peerTLS is
@@ -731,6 +734,7 @@ func MustNewMember(t testutil.TB, mcfg MemberConfig) *Member {
 	if mcfg.ExperimentalMaxLearners != 0 {
 		m.ExperimentalMaxLearners = mcfg.ExperimentalMaxLearners
 	}
+	m.Metrics = mcfg.Metrics
 	m.V2Deprecation = config.V2_DEPR_DEFAULT
 	m.GRPCServerRecorder = &grpctesting.GRPCRecorder{}