Change default OTLP port number

This implements specification change open-telemetry/opentelemetry-specification#1221

To make transition to new port numbers less painful OTLP receiver will
also accept data on the legacy port numbers when it is configured to
use the default endpoint. Users who use the default Collector config
can continue sending data to the legacy ports and have a graceful period
to update their senders to start sending to the new ports.

Note that OTLP/HTTP continues using a separate port number from OTLP/gRPC.
There is separate work in progress to use one port for both.

CHANGELOG.md

@@ -6,6 +6,11 @@
 
 - Remove legacy metrics, they were marked as legacy for ~12 months #2105
 
+## 💡 Enhancements 💡
+
+- Change default OTLP port numbers to 4317/4318, also continue receiving on legacy ports
+  55680/55681 during transition period (#2104).
+
 ## v0.14.0 Beta
 
 ## 🚀 New components 🚀

receiver/otlpreceiver/README.md

@@ -25,9 +25,9 @@ receivers:
 
 The following settings are configurable:
 
-- `endpoint` (default = 0.0.0.0:55680): host:port to which the receiver is
-  going to receive data. The valid syntax is described at
-  https://github.com/grpc/grpc/blob/master/doc/naming.md.
+- `endpoint` (default = 0.0.0.0:4317 for grpc protocol, 0.0.0.0:4318 http protocol):
+  host:port to which the receiver is going to receive data. The valid syntax is
+  described at https://github.com/grpc/grpc/blob/master/doc/naming.md.
 
 ## Advanced Configuration
 
@@ -48,7 +48,7 @@ IMPORTANT: bytes fields are encoded as base64 strings.
 
 To write traces with HTTP/JSON, `POST` to `[address]/v1/traces` for traces,
 to `[address]/v1/metrics` for metrics, to `[address]/v1/logs` for logs. The default
-port is `55681`.
+port is `4318`.
 
 The HTTP/JSON endpoint can also optionally configure
 [CORS](https://fetch.spec.whatwg.org/#cors-protocol), which is enabled by
@@ -59,7 +59,7 @@ receivers:
   otlp:
     protocols:
       http:
-        endpoint: "localhost:55681"
+        endpoint: "localhost:4318"
         cors_allowed_origins:
         - http://test.com
         # Origins can have wildcards with *, use * by itself to match any origin.

receiver/otlpreceiver/config_test.go

@@ -82,7 +82,7 @@ func TestLoadConfig(t *testing.T) {
 			Protocols: Protocols{
 				GRPC: &configgrpc.GRPCServerSettings{
 					NetAddr: confignet.NetAddr{
-						Endpoint:  "0.0.0.0:55680",
+						Endpoint:  "0.0.0.0:4317",
 						Transport: "tcp",
 					},
 					ReadBufferSize: 512 * 1024,
@@ -112,7 +112,7 @@ func TestLoadConfig(t *testing.T) {
 			Protocols: Protocols{
 				GRPC: &configgrpc.GRPCServerSettings{
 					NetAddr: confignet.NetAddr{
-						Endpoint:  "0.0.0.0:55680",
+						Endpoint:  "0.0.0.0:4317",
 						Transport: "tcp",
 					},
 					MaxRecvMsgSizeMiB:    32,
@@ -139,7 +139,7 @@ func TestLoadConfig(t *testing.T) {
 			Protocols: Protocols{
 				GRPC: &configgrpc.GRPCServerSettings{
 					NetAddr: confignet.NetAddr{
-						Endpoint:  "0.0.0.0:55680",
+						Endpoint:  "0.0.0.0:4317",
 						Transport: "tcp",
 					},
 					TLSSetting: &configtls.TLSServerSetting{
@@ -151,7 +151,7 @@ func TestLoadConfig(t *testing.T) {
 					ReadBufferSize: 512 * 1024,
 				},
 				HTTP: &confighttp.HTTPServerSettings{
-					Endpoint: "0.0.0.0:55681",
+					Endpoint: "0.0.0.0:4318",
 					TLSSetting: &configtls.TLSServerSetting{
 						TLSSetting: configtls.TLSSetting{
 							CertFile: "test.crt",
@@ -170,7 +170,7 @@ func TestLoadConfig(t *testing.T) {
 			},
 			Protocols: Protocols{
 				HTTP: &confighttp.HTTPServerSettings{
-					Endpoint:    "0.0.0.0:55681",
+					Endpoint:    "0.0.0.0:4318",
 					CorsOrigins: []string{"https://*.test.com", "https://test.com"},
 				},
 			},

receiver/otlpreceiver/factory.go

@@ -19,6 +19,7 @@ import (
 	"fmt"
 
 	"github.com/spf13/viper"
+	"go.uber.org/zap"
 
 	"go.opentelemetry.io/collector/component"
 	"go.opentelemetry.io/collector/config/configgrpc"
@@ -37,6 +38,11 @@ const (
 	protoGRPC          = "grpc"
 	protoHTTP          = "http"
 	protocolsFieldName = "protocols"
+
+	defaultGRPCEndpoint = "0.0.0.0:4317"
+	defaultHTTPEndpoint = "0.0.0.0:4318"
+	legacyGRPCEndpoint  = "0.0.0.0:55680"
+	legacyHTTPEndpoint  = "0.0.0.0:55681"
 )
 
 func NewFactory() component.ReceiverFactory {
@@ -59,14 +65,14 @@ func createDefaultConfig() configmodels.Receiver {
 		Protocols: Protocols{
 			GRPC: &configgrpc.GRPCServerSettings{
 				NetAddr: confignet.NetAddr{
-					Endpoint:  "0.0.0.0:55680",
+					Endpoint:  defaultGRPCEndpoint,
 					Transport: "tcp",
 				},
 				// We almost write 0 bytes, so no need to tune WriteBufferSize.
 				ReadBufferSize: 512 * 1024,
 			},
 			HTTP: &confighttp.HTTPServerSettings{
-				Endpoint: "0.0.0.0:55681",
+				Endpoint: defaultHTTPEndpoint,
 			},
 		},
 	}
@@ -117,11 +123,11 @@ func customUnmarshaler(componentViperSection *viper.Viper, intoCfg interface{})
 // CreateTracesReceiver creates a  trace receiver based on provided config.
 func createTraceReceiver(
 	ctx context.Context,
-	_ component.ReceiverCreateParams,
+	params component.ReceiverCreateParams,
 	cfg configmodels.Receiver,
 	nextConsumer consumer.TracesConsumer,
 ) (component.TracesReceiver, error) {
-	r, err := createReceiver(cfg)
+	r, err := createReceiver(cfg, params.Logger)
 	if err != nil {
 		return nil, err
 	}
@@ -134,11 +140,11 @@ func createTraceReceiver(
 // CreateMetricsReceiver creates a metrics receiver based on provided config.
 func createMetricsReceiver(
 	ctx context.Context,
-	_ component.ReceiverCreateParams,
+	params component.ReceiverCreateParams,
 	cfg configmodels.Receiver,
 	consumer consumer.MetricsConsumer,
 ) (component.MetricsReceiver, error) {
-	r, err := createReceiver(cfg)
+	r, err := createReceiver(cfg, params.Logger)
 	if err != nil {
 		return nil, err
 	}
@@ -151,11 +157,11 @@ func createMetricsReceiver(
 // CreateLogReceiver creates a log receiver based on provided config.
 func createLogReceiver(
 	ctx context.Context,
-	_ component.ReceiverCreateParams,
+	params component.ReceiverCreateParams,
 	cfg configmodels.Receiver,
 	consumer consumer.LogsConsumer,
 ) (component.LogsReceiver, error) {
-	r, err := createReceiver(cfg)
+	r, err := createReceiver(cfg, params.Logger)
 	if err != nil {
 		return nil, err
 	}
@@ -165,7 +171,7 @@ func createLogReceiver(
 	return r, nil
 }
 
-func createReceiver(cfg configmodels.Receiver) (*otlpReceiver, error) {
+func createReceiver(cfg configmodels.Receiver, logger *zap.Logger) (*otlpReceiver, error) {
 	rCfg := cfg.(*Config)
 
 	// There must be one receiver for both metrics and traces. We maintain a map of
@@ -176,7 +182,7 @@ func createReceiver(cfg configmodels.Receiver) (*otlpReceiver, error) {
 	if !ok {
 		var err error
 		// We don't have a receiver, so create one.
-		receiver, err = newOtlpReceiver(rCfg)
+		receiver, err = newOtlpReceiver(rCfg, logger)
 		if err != nil {
 			return nil, err
 		}

receiver/otlpreceiver/otlp.go

@@ -22,10 +22,12 @@ import (
 	"sync"
 
 	gatewayruntime "github.com/grpc-ecosystem/grpc-gateway/runtime"
+	"go.uber.org/zap"
 	"google.golang.org/grpc"
 
 	"go.opentelemetry.io/collector/component"
 	"go.opentelemetry.io/collector/component/componenterror"
+	"go.opentelemetry.io/collector/config/configgrpc"
 	"go.opentelemetry.io/collector/config/confighttp"
 	"go.opentelemetry.io/collector/consumer"
 	collectorlog "go.opentelemetry.io/collector/internal/data/opentelemetry-proto-gen/collector/logs/v1"
@@ -49,14 +51,17 @@ type otlpReceiver struct {
 
 	stopOnce        sync.Once
 	startServerOnce sync.Once
+
+	logger *zap.Logger
 }
 
 // newOtlpReceiver just creates the OpenTelemetry receiver services. It is the caller's
 // responsibility to invoke the respective Start*Reception methods as well
 // as the various Stop*Reception methods to end it.
-func newOtlpReceiver(cfg *Config) (*otlpReceiver, error) {
+func newOtlpReceiver(cfg *Config, logger *zap.Logger) (*otlpReceiver, error) {
 	r := &otlpReceiver{
-		cfg: cfg,
+		cfg:    cfg,
+		logger: logger,
 	}
 	if cfg.GRPC != nil {
 		opts, err := cfg.GRPC.ToServerOption()
@@ -84,43 +89,89 @@ func newOtlpReceiver(cfg *Config) (*otlpReceiver, error) {
 	return r, nil
 }
 
-// Start runs the trace receiver on the gRPC server. Currently
-// it also enables the metrics receiver too.
-func (r *otlpReceiver) Start(_ context.Context, host component.Host) error {
-	if r.traceReceiver == nil && r.metricsReceiver == nil && r.logReceiver == nil {
-		return errors.New("cannot start receiver: no consumers were specified")
+func (r *otlpReceiver) startGRPCServer(cfg *configgrpc.GRPCServerSettings, host component.Host) error {
+	r.logger.Info("Starting GRPC server on endpoint " + cfg.NetAddr.Endpoint)
+	var gln net.Listener
+	gln, err := cfg.ToListener()
+	if err != nil {
+		return err
 	}
+	go func() {
+		if errGrpc := r.serverGRPC.Serve(gln); errGrpc != nil {
+			host.ReportFatalError(errGrpc)
+		}
+	}()
+	return nil
+}
 
+func (r *otlpReceiver) startHTTPServer(cfg *confighttp.HTTPServerSettings, host component.Host) error {
+	r.logger.Info("Starting HTTP server on endpoint " + cfg.Endpoint)
+	var hln net.Listener
+	hln, err := r.cfg.HTTP.ToListener()
+	if err != nil {
+		return err
+	}
+	go func() {
+		if errHTTP := r.serverHTTP.Serve(hln); errHTTP != nil {
+			host.ReportFatalError(errHTTP)
+		}
+	}()
+	return nil
+}
+
+func (r *otlpReceiver) startProtocolServers(host component.Host) error {
 	var err error
-	r.startServerOnce.Do(func() {
-		if r.cfg.GRPC != nil {
-			var gln net.Listener
-			gln, err = r.cfg.GRPC.ToListener()
+	if r.cfg.GRPC != nil {
+		err = r.startGRPCServer(r.cfg.GRPC, host)
+		if err != nil {
+			return err
+		}
+		if r.cfg.GRPC.NetAddr.Endpoint == defaultGRPCEndpoint {
+			r.logger.Info("Setting up a second GRPC listener on legacy endpoint " + legacyGRPCEndpoint)
+
+			// Copy the config.
+			cfgLegacyGRPC := r.cfg.GRPC
+			// And use the legacy endpoint.
+			cfgLegacyGRPC.NetAddr.Endpoint = legacyGRPCEndpoint
+			err = r.startGRPCServer(cfgLegacyGRPC, host)
 			if err != nil {
-				return
+				return err
 			}
-			go func() {
-				if errGrpc := r.serverGRPC.Serve(gln); errGrpc != nil {
-					host.ReportFatalError(errGrpc)
-				}
-			}()
 		}
-		if r.cfg.HTTP != nil {
-			r.serverHTTP = r.cfg.HTTP.ToServer(
-				r.gatewayMux,
-				confighttp.WithErrorHandler(errorHandler),
-			)
-			var hln net.Listener
-			hln, err = r.cfg.HTTP.ToListener()
+	}
+	if r.cfg.HTTP != nil {
+		r.serverHTTP = r.cfg.HTTP.ToServer(
+			r.gatewayMux,
+			confighttp.WithErrorHandler(errorHandler),
+		)
+		err = r.startHTTPServer(r.cfg.HTTP, host)
+		if err != nil {
+			return err
+		}
+		if r.cfg.HTTP.Endpoint == defaultHTTPEndpoint {
+			r.logger.Info("Setting up a second HTTP listener on legacy endpoint " + legacyHTTPEndpoint)
+			cfgLegacyHTTP := r.cfg.HTTP
+			cfgLegacyHTTP.Endpoint = legacyHTTPEndpoint
+			err = r.startHTTPServer(cfgLegacyHTTP, host)
 			if err != nil {
-				return
+				return err
 			}
-			go func() {
-				if errHTTP := r.serverHTTP.Serve(hln); errHTTP != nil {
-					host.ReportFatalError(errHTTP)
-				}
-			}()
 		}
+	}
+
+	return err
+}
+
+// Start runs the trace receiver on the gRPC server. Currently
+// it also enables the metrics receiver too.
+func (r *otlpReceiver) Start(_ context.Context, host component.Host) error {
+	if r.traceReceiver == nil && r.metricsReceiver == nil && r.logReceiver == nil {
+		return errors.New("cannot start receiver: no consumers were specified")
+	}
+
+	var err error
+	r.startServerOnce.Do(func() {
+		err = r.startProtocolServers(host)
 	})
 	return err
 }

receiver/otlpreceiver/otlp_test.go

@@ -30,6 +30,7 @@ import (
 	"github.com/gogo/protobuf/jsonpb"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"go.uber.org/zap"
 	spb "google.golang.org/genproto/googleapis/rpc/status"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/codes"
@@ -698,7 +699,7 @@ func TestGRPCInvalidTLSCredentials(t *testing.T) {
 	}
 
 	// TLS is resolved during Creation of the receiver for GRPC.
-	_, err := createReceiver(cfg)
+	_, err := createReceiver(cfg, zap.NewNop())
 	assert.EqualError(t, err,
 		`failed to load TLS config: for auth via TLS, either both certificate and key must be supplied, or neither`)
 }
@@ -745,7 +746,7 @@ func newHTTPReceiver(t *testing.T, endpoint string, tc consumer.TracesConsumer,
 }
 
 func newReceiver(t *testing.T, factory component.ReceiverFactory, cfg *Config, tc consumer.TracesConsumer, mc consumer.MetricsConsumer) *otlpReceiver {
-	r, err := createReceiver(cfg)
+	r, err := createReceiver(cfg, zap.NewNop())
 	require.NoError(t, err)
 	if tc != nil {
 		params := component.ReceiverCreateParams{}