tracing fixes

Signed-off-by: Vasiliy Tolstov <v.tolstov@unistack.org>

event.go

@@ -1,12 +1,14 @@
 package kgo
 
 import (
+	"context"
 	"sync"
 
 	"go.unistack.org/micro/v3/broker"
 )
 
 type event struct {
+	ctx   context.Context
 	topic string
 	err   error
 	sync.RWMutex
@@ -14,6 +16,10 @@ type event struct {
 	ack bool
 }
 
+func (p *event) Context() context.Context {
+	return p.ctx
+}
+
 func (p *event) Topic() string {
 	return p.topic
 }

go.mod

@@ -1,21 +1,24 @@
 module go.unistack.org/micro-broker-kgo/v3
 
-go 1.19
+go 1.21
+
+toolchain go1.22.1
 
 require (
 	github.com/google/uuid v1.6.0
 	github.com/twmb/franz-go v1.16.1
+	github.com/twmb/franz-go/pkg/kadm v1.11.0
 	github.com/twmb/franz-go/pkg/kmsg v1.7.0
-	go.opentelemetry.io/otel v1.24.0
-	go.unistack.org/micro/v3 v3.10.46
+	go.opentelemetry.io/otel v1.25.0
+	go.unistack.org/micro/v3 v3.10.59
 )
 
 require (
-	github.com/golang/protobuf v1.5.4 // indirect
-	github.com/klauspost/compress v1.17.7 // indirect
+	github.com/klauspost/compress v1.17.8 // indirect
 	github.com/pierrec/lz4/v4 v4.1.21 // indirect
-	golang.org/x/sys v0.18.0 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20240304212257-790db918fca8 // indirect
-	google.golang.org/grpc v1.62.1 // indirect
+	golang.org/x/crypto v0.22.0 // indirect
+	golang.org/x/sys v0.19.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56 // indirect
+	google.golang.org/grpc v1.63.2 // indirect
 	google.golang.org/protobuf v1.33.0 // indirect
 )

go.sum

@@ -1,32 +1,40 @@
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
-github.com/golang/protobuf v1.5.4 h1:i7eJL8qZTpSEXOPTxNKhASYpMn+8e5Q6AdndVa1dWek=
-github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6rSs7xps=
+github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
+github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
-github.com/klauspost/compress v1.17.7 h1:ehO88t2UGzQK66LMdE8tibEd1ErmzZjNEqWkjLAKQQg=
-github.com/klauspost/compress v1.17.7/go.mod h1:Di0epgTjJY877eYKx5yC51cX2A2Vl2ibi7bDH9ttBbw=
+github.com/klauspost/compress v1.17.8 h1:YcnTYrq7MikUT7k0Yb5eceMmALQPYBW/Xltxn0NAMnU=
+github.com/klauspost/compress v1.17.8/go.mod h1:Di0epgTjJY877eYKx5yC51cX2A2Vl2ibi7bDH9ttBbw=
 github.com/pierrec/lz4/v4 v4.1.21 h1:yOVMLb6qSIDP67pl/5F7RepeKYu/VmTyEXvuMI5d9mQ=
 github.com/pierrec/lz4/v4 v4.1.21/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
-github.com/stretchr/testify v1.8.4 h1:CcVxjf3Q8PM0mHUKJCdn+eZZtm5yQwehR5yeSVQQcUk=
+github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
+github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg=
+github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
 github.com/twmb/franz-go v1.16.1 h1:rpWc7fB9jd7TgmCyfxzenBI+QbgS8ZfJOUQE+tzPtbE=
 github.com/twmb/franz-go v1.16.1/go.mod h1:/pER254UPPGp/4WfGqRi+SIRGE50RSQzVubQp6+N4FA=
+github.com/twmb/franz-go/pkg/kadm v1.11.0 h1:FfeWJ0qadntFpAcQt8JzNXW4dijjytZNLrzJuzzzuxA=
+github.com/twmb/franz-go/pkg/kadm v1.11.0/go.mod h1:qrhkdH+SWS3ivmbqOgHbpgVHamhaKcjH0UM+uOp0M1A=
 github.com/twmb/franz-go/pkg/kmsg v1.7.0 h1:a457IbvezYfA5UkiBvyV3zj0Is3y1i8EJgqjJYoij2E=
 github.com/twmb/franz-go/pkg/kmsg v1.7.0/go.mod h1:se9Mjdt0Nwzc9lnjJ0HyDtLyBnaBDAd7pCje47OhSyw=
-go.opentelemetry.io/otel v1.24.0 h1:0LAOdjNmQeSTzGBzduGe/rU4tZhMwL5rWgtp9Ku5Jfo=
-go.opentelemetry.io/otel v1.24.0/go.mod h1:W7b9Ozg4nkF5tWI5zsXkaKKDjdVjpD4oAt9Qi/MArHo=
-go.unistack.org/micro/v3 v3.10.46 h1:rnuEqiFkerwJmKzHmHBXRgxFemZustxWz2hRNLQQ8cU=
-go.unistack.org/micro/v3 v3.10.46/go.mod h1:erMgt3Bl7vQQ0e9UpQyR5NlLiZ9pKeEJ9+1tfYFaqUg=
-golang.org/x/crypto v0.17.0 h1:r8bRNjWL3GshPW3gkd+RpvzWrZAwPS49OmTGZ/uhM4k=
-golang.org/x/net v0.20.0 h1:aCL9BSgETF1k+blQaYUBx9hJ9LOGP3gAVemcZlf1Kpo=
-golang.org/x/sys v0.18.0 h1:DBdB3niSjOA/O0blCZBqDefyWNYveAYMNF1Wum0DYQ4=
-golang.org/x/sys v0.18.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+go.opentelemetry.io/otel v1.25.0 h1:gldB5FfhRl7OJQbUHt/8s0a7cE8fbsPAtdpRaApKy4k=
+go.opentelemetry.io/otel v1.25.0/go.mod h1:Wa2ds5NOXEMkCmUou1WA7ZBfLTHWIsp034OVD7AO+Vg=
+go.unistack.org/micro/v3 v3.10.59 h1:eneYXJLgyu5MZpSvyI0K17CeXvgOoUCN5dWZaPV5lI4=
+go.unistack.org/micro/v3 v3.10.59/go.mod h1:erMgt3Bl7vQQ0e9UpQyR5NlLiZ9pKeEJ9+1tfYFaqUg=
+golang.org/x/crypto v0.22.0 h1:g1v0xeRhjcugydODzvb3mEM9SQ0HGp9s/nh3COQ/C30=
+golang.org/x/crypto v0.22.0/go.mod h1:vr6Su+7cTlO45qkww3VDJlzDn0ctJvRgYbC2NvXHt+M=
+golang.org/x/net v0.21.0 h1:AQyQV4dYCvJ7vGmJyKki9+PBdyvhkSd8EIx/qb0AYv4=
+golang.org/x/net v0.21.0/go.mod h1:bIjVDfnllIU7BJ2DNgfnXvpSvtn8VRwhlsaeUTyUS44=
+golang.org/x/sys v0.19.0 h1:q5f1RH2jigJ1MoAWp2KTp3gm5zAGFUTarQZ5U386+4o=
+golang.org/x/sys v0.19.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
 golang.org/x/text v0.14.0 h1:ScX5w1eTa3QqT8oi6+ziP7dTV1S2+ALU0bI+0zXKWiQ=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20240304212257-790db918fca8 h1:IR+hp6ypxjH24bkMfEJ0yHR21+gwPWdV+/IBrPQyn3k=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20240304212257-790db918fca8/go.mod h1:UCOku4NytXMJuLQE5VuqA5lX3PcHCBo8pxNyvkf4xBs=
-google.golang.org/grpc v1.62.1 h1:B4n+nfKzOICUXMgyrNd19h/I9oH0L1pizfk1d4zSgTk=
-google.golang.org/grpc v1.62.1/go.mod h1:IWTG0VlJLCh1SkC58F7np9ka9mx/WNkjl4PGJaiq+QE=
+golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56 h1:zviK8GX4VlMstrK3JkexM5UHjH1VOkRebH9y3jhSBGk=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56/go.mod h1:WtryC6hu0hhx87FDGxWCDptyssuo68sk10vYjF+T9fY=
+google.golang.org/grpc v1.63.2 h1:MUeiw1B2maTVZthpU5xvASfTh3LDbxHd6IJ6QQVU+xM=
+google.golang.org/grpc v1.63.2/go.mod h1:WAX/8DgncnokcFUldAxq7GeB5DXHDbMF+lLvDomNkRA=
 google.golang.org/protobuf v1.33.0 h1:uNO2rsAINq/JlFpSdYEKIZ0uKD/R9cpdv0T+yoGwGmI=
 google.golang.org/protobuf v1.33.0/go.mod h1:c6P6GXX6sHbq/GpV6MGZEdwhWPcYBgnhAHhKbcUYpos=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
+gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

kgo.go

@@ -14,7 +14,9 @@ import (
 	"github.com/twmb/franz-go/pkg/kgo"
 	"github.com/twmb/franz-go/pkg/kmsg"
 	"go.unistack.org/micro/v3/broker"
+	"go.unistack.org/micro/v3/logger"
 	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/semconv"
 	"go.unistack.org/micro/v3/tracer"
 	mrand "go.unistack.org/micro/v3/util/rand"
 )
@@ -60,7 +62,7 @@ type Broker struct {
 	connected bool
 	sync.RWMutex
 	opts broker.Options
-	subs []*subscriber
+	subs []*Subscriber
 }
 
 func (k *Broker) Address() string {
@@ -71,12 +73,15 @@ func (k *Broker) Name() string {
 	return k.opts.Name
 }
 
-func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, error) {
+func (k *Broker) Client() *kgo.Client {
+	return k.c
+}
+
+func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, *hookTracer, error) {
 	var c *kgo.Client
 	var err error
-	var span tracer.Span
-	ctx, span = k.opts.Tracer.Start(ctx, "Connect")
-	defer span.Finish()
+
+	sp, _ := tracer.SpanFromContext(ctx)
 
 	clientID := "kgo"
 	group := ""
@@ -89,28 +94,33 @@ func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, err
 		}
 	}
 
+	htracer := &hookTracer{group: group, clientID: clientID, tracer: k.opts.Tracer}
 	opts = append(opts,
 		kgo.WithHooks(&hookMeter{meter: k.opts.Meter}),
-		kgo.WithHooks(&hookTracer{group: group, clientID: clientID, tracer: k.opts.Tracer}),
+		kgo.WithHooks(htracer),
 	)
 
 	select {
 	case <-ctx.Done():
 		if ctx.Err() != nil {
-			span.SetStatus(tracer.SpanStatusError, ctx.Err().Error())
+			if sp != nil {
+				sp.SetStatus(tracer.SpanStatusError, ctx.Err().Error())
+			}
 		}
-		return nil, ctx.Err()
+		return nil, nil, ctx.Err()
 	default:
 		c, err = kgo.NewClient(opts...)
 		if err == nil {
 			err = c.Ping(ctx) // check connectivity to cluster
 		}
 		if err != nil {
-			span.SetStatus(tracer.SpanStatusError, err.Error())
-			return nil, err
+			if sp != nil {
+				sp.SetStatus(tracer.SpanStatusError, err.Error())
+			}
+			return nil, nil, err
 		}
 	}
-	return c, nil
+	return c, htracer, nil
 }
 
 func (k *Broker) Connect(ctx context.Context) error {
@@ -126,7 +136,7 @@ func (k *Broker) Connect(ctx context.Context) error {
 		nctx = ctx
 	}
 
-	c, err := k.connect(nctx, k.kopts...)
+	c, _, err := k.connect(nctx, k.kopts...)
 	if err != nil {
 		return err
 	}
@@ -229,13 +239,9 @@ func (k *Broker) Publish(ctx context.Context, topic string, msg *broker.Message,
 }
 
 func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...broker.PublishOption) error {
-	var span tracer.Span
-	ctx, span = k.opts.Tracer.Start(ctx, "Publish")
-	defer span.Finish()
-
 	k.Lock()
 	if !k.connected {
-		c, err := k.connect(ctx, k.kopts...)
+		c, _, err := k.connect(ctx, k.kopts...)
 		if err != nil {
 			k.Unlock()
 			return err
@@ -265,7 +271,7 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 		rec := &kgo.Record{Context: ctx, Key: key}
 		rec.Topic, _ = msg.Header.Get(metadata.HeaderTopic)
 		msg.Header.Del(metadata.HeaderTopic)
-		// k.opts.Meter.Counter(broker.PublishMessageInflight, "endpoint", rec.Topic).Inc()
+		k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Inc()
 		if options.BodyOnly || k.opts.Codec.String() == "noop" {
 			rec.Value = msg.Body
 			for k, v := range msg.Header {
@@ -281,35 +287,35 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 	}
 
 	if promise != nil {
-		// ts := time.Now()
+		ts := time.Now()
 		for _, rec := range records {
 			k.c.Produce(ctx, rec, func(r *kgo.Record, err error) {
-				// te := time.Since(ts)
-				//	k.opts.Meter.Counter(broker.PublishMessageInflight, "endpoint", rec.Topic).Dec()
-				//	k.opts.Meter.Summary(broker.PublishMessageLatencyMicroseconds, "endpoint", r.Topic).Update(te.Seconds())
-				//	k.opts.Meter.Histogram(broker.PublishMessageDurationSeconds, "endpoint", r.Topic).Update(te.Seconds())
+				te := time.Since(ts)
+				k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Dec()
+				k.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", rec.Topic, "topic", rec.Topic).Update(te.Seconds())
+				k.opts.Meter.Histogram(semconv.PublishMessageDurationSeconds, "endpoint", rec.Topic, "topic", rec.Topic).Update(te.Seconds())
 				if err != nil {
-					//	k.opts.Meter.Counter(broker.PublishMessageTotal, "endpoint", r.Topic, "status", "failure").Inc()
+					k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "failure").Inc()
 				} else {
-					// k.opts.Meter.Counter(broker.PublishMessageTotal, "endpoint", r.Topic, "status", "success").Inc()
+					k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "success").Inc()
 				}
 				promise(r, err)
 			})
 		}
 		return nil
 	}
-	// ts := time.Now()
+	ts := time.Now()
 	results := k.c.ProduceSync(ctx, records...)
-	// te := time.Since(ts)
+	te := time.Since(ts)
 	for _, result := range results {
-		//	k.opts.Meter.Summary(broker.PublishMessageLatencyMicroseconds, "endpoint", result.Record.Topic).Update(te.Seconds())
-		//	k.opts.Meter.Histogram(broker.PublishMessageDurationSeconds, "endpoint", result.Record.Topic).Update(te.Seconds())
-		///		k.opts.Meter.Counter(broker.PublishMessageInflight, "endpoint", result.Record.Topic).Dec()
+		k.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+		k.opts.Meter.Histogram(semconv.PublishMessageDurationSeconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+		k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Dec()
 		if result.Err != nil {
-			//			k.opts.Meter.Counter(broker.PublishMessageTotal, "endpoint", result.Record.Topic, "status", "failure").Inc()
+			k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "failure").Inc()
 			errs = append(errs, result.Err.Error())
 		} else {
-			//			k.opts.Meter.Counter(broker.PublishMessageTotal, "endpoint", result.Record.Topic, "status", "success").Inc()
+			k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "success").Inc()
 		}
 	}
 
@@ -320,6 +326,22 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 	return nil
 }
 
+func (k *Broker) TopicExists(ctx context.Context, topic string) error {
+	mdreq := kmsg.NewMetadataRequest()
+	mdreq.Topics = []kmsg.MetadataRequestTopic{
+		{Topic: &topic},
+	}
+
+	mdrsp, err := mdreq.RequestWith(ctx, k.c)
+	if err != nil {
+		return err
+	} else if mdrsp.Topics[0].ErrorCode != 0 {
+		return fmt.Errorf("topic %s not exists or permission error", topic)
+	}
+
+	return nil
+}
+
 func (k *Broker) BatchSubscribe(ctx context.Context, topic string, handler broker.BatchHandler, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
 	return nil, nil
 }
@@ -342,7 +364,7 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	sub := &subscriber{
+	sub := &Subscriber{
 		topic:     topic,
 		opts:      options,
 		handler:   handler,
@@ -369,7 +391,7 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	c, err := k.connect(ctx, kopts...)
+	c, htracer, err := k.connect(ctx, kopts...)
 	if err != nil {
 		return nil, err
 	}
@@ -387,6 +409,8 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	}
 
 	sub.c = c
+	sub.htracer = htracer
+
 	go sub.poll(ctx)
 
 	k.Lock()
@@ -412,7 +436,7 @@ func NewBroker(opts ...broker.Option) *Broker {
 		kgo.DialTimeout(3 * time.Second),
 		kgo.DisableIdempotentWrite(),
 		kgo.ProducerBatchCompression(kgo.NoCompression()),
-		kgo.WithLogger(&mlogger{l: options.Logger, ctx: options.Context}),
+		kgo.WithLogger(&mlogger{l: options.Logger.Clone(logger.WithCallerSkipCount(options.Logger.Options().CallerSkipCount + 2)), ctx: options.Context}),
 		kgo.SeedBrokers(kaddrs...),
 		kgo.RetryBackoffFn(DefaultRetryBackoffFn),
 		kgo.BlockRebalanceOnPoll(),

meter.go

@@ -14,14 +14,21 @@ type hookMeter struct {
 }
 
 var (
-	_ kgo.HookBrokerConnect       = &hookMeter{}
-	_ kgo.HookBrokerDisconnect    = &hookMeter{}
-	_ kgo.HookBrokerRead          = &hookMeter{}
-	_ kgo.HookBrokerThrottle      = &hookMeter{}
-	_ kgo.HookBrokerWrite         = &hookMeter{}
-	_ kgo.HookFetchBatchRead      = &hookMeter{}
+	_ kgo.HookBrokerConnect    = &hookMeter{}
+	_ kgo.HookBrokerDisconnect = &hookMeter{}
+	// HookBrokerE2E
+	_ kgo.HookBrokerRead     = &hookMeter{}
+	_ kgo.HookBrokerThrottle = &hookMeter{}
+	_ kgo.HookBrokerWrite    = &hookMeter{}
+	_ kgo.HookFetchBatchRead = &hookMeter{}
+	// HookFetchRecordBuffered
+	// HookFetchRecordUnbuffered
+	_ kgo.HookGroupManageError = &hookMeter{}
+	// HookNewClient
 	_ kgo.HookProduceBatchWritten = &hookMeter{}
-	_ kgo.HookGroupManageError    = &hookMeter{}
+	// HookProduceRecordBuffered
+	// HookProduceRecordPartitioned
+	// HookProduceRecordUnbuffered
 )
 
 const (

options.go

@@ -9,8 +9,17 @@ import (
 	"go.unistack.org/micro/v3/client"
 )
 
-// DefaultCommitInterval specifies how fast send commit offsets to kafka
-var DefaultCommitInterval = 5 * time.Second
+var (
+
+	// DefaultCommitInterval specifies how fast send commit offsets to kafka
+	DefaultCommitInterval = 5 * time.Second
+
+	// DefaultStatsInterval specifies how fast check consumer lag
+	DefaultStatsInterval = 5 * time.Second
+
+	// DefaultSubscribeMaxInflight specifies how much messages keep inflight
+	DefaultSubscribeMaxInflight = 100
+)
 
 type subscribeContextKey struct{}
 
@@ -82,8 +91,6 @@ func CommitInterval(td time.Duration) broker.Option {
 	return broker.SetOption(commitIntervalKey{}, td)
 }
 
-var DefaultSubscribeMaxInflight = 10
-
 type subscribeMaxInflightKey struct{}
 
 // SubscribeMaxInFlight max queued messages

subscriber.go

@@ -2,12 +2,17 @@ package kgo
 
 import (
 	"context"
+	"strconv"
 	"sync"
+	"time"
 
+	"github.com/twmb/franz-go/pkg/kadm"
 	"github.com/twmb/franz-go/pkg/kgo"
 	"go.unistack.org/micro/v3/broker"
 	"go.unistack.org/micro/v3/logger"
 	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/semconv"
+	"go.unistack.org/micro/v3/tracer"
 )
 
 type tp struct {
@@ -19,6 +24,7 @@ type consumer struct {
 	c         *kgo.Client
 	topic     string
 	partition int32
+	htracer   *hookTracer
 	opts      broker.SubscribeOptions
 	kopts     broker.Options
 	handler   broker.Handler
@@ -27,9 +33,10 @@ type consumer struct {
 	recs      chan kgo.FetchTopicPartition
 }
 
-type subscriber struct {
+type Subscriber struct {
 	c         *kgo.Client
 	topic     string
+	htracer   *hookTracer
 	opts      broker.SubscribeOptions
 	kopts     broker.Options
 	handler   broker.Handler
@@ -39,15 +46,19 @@ type subscriber struct {
 	sync.RWMutex
 }
 
-func (s *subscriber) Options() broker.SubscribeOptions {
+func (s *Subscriber) Client() *kgo.Client {
+	return s.c
+}
+
+func (s *Subscriber) Options() broker.SubscribeOptions {
 	return s.opts
 }
 
-func (s *subscriber) Topic() string {
+func (s *Subscriber) Topic() string {
 	return s.topic
 }
 
-func (s *subscriber) Unsubscribe(ctx context.Context) error {
+func (s *Subscriber) Unsubscribe(ctx context.Context) error {
 	if s.closed {
 		return nil
 	}
@@ -69,13 +80,48 @@ func (s *subscriber) Unsubscribe(ctx context.Context) error {
 	return nil
 }
 
-func (s *subscriber) poll(ctx context.Context) {
+func (s *Subscriber) poll(ctx context.Context) {
 	maxInflight := DefaultSubscribeMaxInflight
 	if s.opts.Context != nil {
 		if n, ok := s.opts.Context.Value(subscribeMaxInflightKey{}).(int); n > 0 && ok {
 			maxInflight = n
 		}
 	}
+
+	go func() {
+		ac := kadm.NewClient(s.c)
+		ticker := time.NewTicker(DefaultStatsInterval)
+
+		for {
+			select {
+			case <-ctx.Done():
+				ticker.Stop()
+				return
+			case <-ticker.C:
+				dgls, err := ac.Lag(ctx, s.opts.Group)
+				if err != nil || !dgls.Ok() {
+					continue
+				}
+
+				dgl, ok := dgls[s.opts.Group]
+				if !ok {
+					continue
+				}
+				lmap, ok := dgl.Lag[s.topic]
+				if !ok {
+					continue
+				}
+
+				s.Lock()
+				for p, l := range lmap {
+					s.kopts.Meter.Counter(semconv.BrokerGroupLag, "topic", s.topic, "group", s.opts.Group, "partition", strconv.Itoa(int(p)), "lag", strconv.Itoa(int(l.Lag)))
+				}
+				s.Unlock()
+
+			}
+		}
+	}()
+
 	for {
 		select {
 		case <-ctx.Done():
@@ -102,7 +148,7 @@ func (s *subscriber) poll(ctx context.Context) {
 	}
 }
 
-func (s *subscriber) killConsumers(ctx context.Context, lost map[string][]int32) {
+func (s *Subscriber) killConsumers(ctx context.Context, lost map[string][]int32) {
 	var wg sync.WaitGroup
 	defer wg.Wait()
 
@@ -119,12 +165,12 @@ func (s *subscriber) killConsumers(ctx context.Context, lost map[string][]int32)
 	}
 }
 
-func (s *subscriber) lost(ctx context.Context, _ *kgo.Client, lost map[string][]int32) {
+func (s *Subscriber) lost(ctx context.Context, _ *kgo.Client, lost map[string][]int32) {
 	s.kopts.Logger.Debugf(ctx, "[kgo] lost %#+v", lost)
 	s.killConsumers(ctx, lost)
 }
 
-func (s *subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[string][]int32) {
+func (s *Subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[string][]int32) {
 	s.kopts.Logger.Debugf(ctx, "[kgo] revoked %#+v", revoked)
 	s.killConsumers(ctx, revoked)
 	if err := c.CommitMarkedOffsets(ctx); err != nil {
@@ -132,22 +178,24 @@ func (s *subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[str
 	}
 }
 
-func (s *subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[string][]int32) {
+func (s *Subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[string][]int32) {
 	for topic, partitions := range assigned {
 		for _, partition := range partitions {
 			pc := &consumer{
 				c:         c,
 				topic:     topic,
 				partition: partition,
-
-				quit:    make(chan struct{}),
-				done:    make(chan struct{}),
-				recs:    make(chan kgo.FetchTopicPartition, 100),
-				handler: s.handler,
-				kopts:   s.kopts,
-				opts:    s.opts,
+				htracer:   s.htracer,
+				quit:      make(chan struct{}),
+				done:      make(chan struct{}),
+				recs:      make(chan kgo.FetchTopicPartition, 100),
+				handler:   s.handler,
+				kopts:     s.kopts,
+				opts:      s.opts,
 			}
+			s.Lock()
 			s.consumers[tp{topic, partition}] = pc
+			s.Unlock()
 			go pc.consume()
 		}
 	}
@@ -169,8 +217,9 @@ func (pc *consumer) consume() {
 			return
 		case p := <-pc.recs:
 			for _, record := range p.Records {
-				// ts := time.Now()
-				//	pc.kopts.Meter.Counter(broker.SubscribeMessageInflight, "endpoint", record.Topic).Inc()
+				ctx, sp := pc.htracer.WithProcessSpan(record)
+				ts := time.Now()
+				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Inc()
 				p := eventPool.Get().(*event)
 				p.msg.Header = nil
 				p.msg.Body = nil
@@ -178,6 +227,7 @@ func (pc *consumer) consume() {
 				p.err = nil
 				p.ack = false
 				p.msg.Header = metadata.New(len(record.Headers))
+				p.ctx = ctx
 				for _, hdr := range record.Headers {
 					p.msg.Header.Set(hdr.Key, string(hdr.Value))
 				}
@@ -186,13 +236,17 @@ func (pc *consumer) consume() {
 				} else if pc.opts.BodyOnly {
 					p.msg.Body = record.Value
 				} else {
-					if err := pc.kopts.Codec.Unmarshal(record.Value, p.msg); err != nil {
-						//	pc.kopts.Meter.Counter(broker.SubscribeMessageTotal, "endpoint", record.Topic, "status", "failure").Inc()
+					sp.AddEvent("codec unmarshal start")
+					err := pc.kopts.Codec.Unmarshal(record.Value, p.msg)
+					sp.AddEvent("codec unmarshal stop")
+					if err != nil {
+						sp.SetStatus(tracer.SpanStatusError, err.Error())
+						pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "failure").Inc()
 						p.err = err
 						p.msg.Body = record.Value
 						if eh != nil {
 							_ = eh(p)
-							//	pc.kopts.Meter.Counter(broker.SubscribeMessageInflight, "endpoint", record.Topic).Dec()
+							pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 							if p.ack {
 								pc.c.MarkCommitRecords(record)
 							} else {
@@ -201,54 +255,63 @@ func (pc *consumer) consume() {
 								return
 							}
 							eventPool.Put(p)
-							// te := time.Since(ts)
-							//	pc.kopts.Meter.Summary(broker.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic).Update(te.Seconds())
-							//	pc.kopts.Meter.Histogram(broker.SubscribeMessageDurationSeconds, "endpoint", record.Topic).Update(te.Seconds())
+							te := time.Since(ts)
+							pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+							pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 							continue
 						} else {
 							if pc.kopts.Logger.V(logger.ErrorLevel) {
 								pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: failed to unmarshal: %v", err)
 							}
 						}
-						//		te := time.Since(ts)
-						//		pc.kopts.Meter.Counter(broker.SubscribeMessageInflight, "endpoint", record.Topic).Dec()
-						//		pc.kopts.Meter.Summary(broker.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic).Update(te.Seconds())
-						//			pc.kopts.Meter.Histogram(broker.SubscribeMessageDurationSeconds, "endpoint", record.Topic).Update(te.Seconds())
+						te := time.Since(ts)
+						pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
+						pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+						pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 						eventPool.Put(p)
 						pc.kopts.Logger.Fatalf(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
+						sp.Finish()
 						return
 					}
 				}
+				sp.AddEvent("handler start")
 				err := pc.handler(p)
+				sp.AddEvent("handler stop")
 				if err == nil {
-					//	pc.kopts.Meter.Counter(broker.SubscribeMessageTotal, "endpoint", record.Topic, "status", "success").Inc()
+					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "success").Inc()
 				} else {
-					//	pc.kopts.Meter.Counter(broker.SubscribeMessageTotal, "endpoint", record.Topic, "status", "failure").Inc()
+					sp.SetStatus(tracer.SpanStatusError, err.Error())
+					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "failure").Inc()
 				}
-				// pc.kopts.Meter.Counter(broker.SubscribeMessageInflight, "endpoint", record.Topic).Dec()
+				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 				if err == nil && pc.opts.AutoAck {
 					p.ack = true
 				} else if err != nil {
 					p.err = err
 					if eh != nil {
+						sp.AddEvent("error handler start")
 						_ = eh(p)
+						sp.AddEvent("error handler stop")
 					} else {
 						if pc.kopts.Logger.V(logger.ErrorLevel) {
 							pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: subscriber error: %v", err)
 						}
 					}
 				}
-				//	te := time.Since(ts)
-				//	pc.kopts.Meter.Summary(broker.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic).Update(te.Seconds())
-				//	pc.kopts.Meter.Histogram(broker.SubscribeMessageDurationSeconds, "endpoint", record.Topic).Update(te.Seconds())
+				te := time.Since(ts)
+				pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+				pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 				if p.ack {
 					eventPool.Put(p)
 					pc.c.MarkCommitRecords(record)
 				} else {
 					eventPool.Put(p)
 					pc.kopts.Logger.Fatalf(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+					sp.SetStatus(tracer.SpanStatusError, "ErrLostMessage")
+					sp.Finish()
 					return
 				}
+				sp.Finish()
 			}
 		}
 	}

tracer.go

@@ -2,14 +2,12 @@ package kgo
 
 import (
 	"context"
-	"net"
-	"time"
 	"unicode/utf8"
 
 	"github.com/twmb/franz-go/pkg/kgo"
-	"go.unistack.org/micro/v3/tracer"
-	"go.opentelemetry.io/otel/attribute"
 	semconv "go.opentelemetry.io/otel/semconv/v1.18.0"
+	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/tracer"
 )
 
 type hookTracer struct {
@@ -19,40 +17,12 @@ type hookTracer struct {
 }
 
 var (
-	_ kgo.HookBrokerConnect       = &hookTracer{}
-	_ kgo.HookBrokerDisconnect    = &hookTracer{}
-	_ kgo.HookBrokerRead          = &hookTracer{}
-	_ kgo.HookBrokerThrottle      = &hookTracer{}
-	_ kgo.HookBrokerWrite         = &hookTracer{}
-	_ kgo.HookFetchBatchRead      = &hookTracer{}
-	_ kgo.HookProduceBatchWritten = &hookTracer{}
-	_ kgo.HookGroupManageError    = &hookTracer{}
+	_ kgo.HookProduceRecordBuffered   = (*hookTracer)(nil)
+	_ kgo.HookProduceRecordUnbuffered = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordBuffered     = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordUnbuffered   = (*hookTracer)(nil)
 )
 
-func (m *hookTracer) OnGroupManageError(err error) {
-}
-
-func (m *hookTracer) OnBrokerConnect(meta kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
-}
-
-func (m *hookTracer) OnBrokerDisconnect(meta kgo.BrokerMetadata, _ net.Conn) {
-}
-
-func (m *hookTracer) OnBrokerWrite(meta kgo.BrokerMetadata, _ int16, bytesWritten int, writeWait, timeToWrite time.Duration, err error) {
-}
-
-func (m *hookTracer) OnBrokerRead(meta kgo.BrokerMetadata, _ int16, bytesRead int, readWait, timeToRead time.Duration, err error) {
-}
-
-func (m *hookTracer) OnBrokerThrottle(meta kgo.BrokerMetadata, throttleInterval time.Duration, _ bool) {
-}
-
-func (m *hookTracer) OnProduceBatchWritten(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.ProduceBatchMetrics) {
-}
-
-func (m *hookTracer) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.FetchBatchMetrics) {
-}
-
 // OnProduceRecordBuffered starts a new span for the "publish" operation on a
 // buffered record.
 //
@@ -61,7 +31,7 @@ func (m *hookTracer) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ i
 // hook.
 func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 	// Set up span options.
-	attrs := []attribute.KeyValue{
+	attrs := []interface{}{
 		semconv.MessagingSystemKey.String("kafka"),
 		semconv.MessagingDestinationKindTopic,
 		semconv.MessagingDestinationName(r.Topic),
@@ -71,20 +41,32 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 	if m.clientID != "" {
 		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
 	}
-	ifattrs := make([]interface{}, 0, len(attrs))
-	for _, attr := range attrs {
-		ifattrs = append(ifattrs, attr)
-	}
 	opts := []tracer.SpanOption{
-		tracer.WithSpanLabels(ifattrs...),
+		tracer.WithSpanLabels(attrs...),
 		tracer.WithSpanKind(tracer.SpanKindProducer),
 	}
-	// Start the "publish" span.
-	ctx, _ := m.tracer.Start(r.Context, r.Topic+" publish", opts...)
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	md, ok := metadata.FromOutgoingContext(r.Context)
+	if !ok {
+		md = metadata.New(len(r.Headers))
+	}
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
 	// Inject the span context into the record.
 	// t.propagators.Inject(ctx, NewRecordCarrier(r))
 	// Update the record context.
-	r.Context = ctx
+
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewOutgoingContext(r.Context, md), r.Topic+" publish", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, r.Topic+" publish", opts...)
+	}
 }
 
 // OnProduceRecordUnbuffered continues and ends the "publish" span for an
@@ -93,17 +75,14 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 // It sets attributes with values unset when producing and records any error
 // that occurred during the publish operation.
 func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
-	span, ok := tracer.SpanFromContext(r.Context)
-	if !ok {
-		return
-	}
-	defer span.Finish()
+	span, _ := tracer.SpanFromContext(r.Context)
 	span.AddLabels(
 		semconv.MessagingKafkaDestinationPartition(int(r.Partition)),
 	)
 	if err != nil {
 		span.SetStatus(tracer.SpanStatusError, err.Error())
 	}
+	span.Finish()
 }
 
 // OnFetchRecordBuffered starts a new span for the "receive" operation on a
@@ -115,7 +94,7 @@ func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
 // processing.
 func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
 	// Set up the span options.
-	attrs := []attribute.KeyValue{
+	attrs := []interface{}{
 		semconv.MessagingSystemKey.String("kafka"),
 		semconv.MessagingSourceKindTopic,
 		semconv.MessagingSourceName(r.Topic),
@@ -129,32 +108,39 @@ func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
 	if m.group != "" {
 		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
 	}
-	ifattrs := make([]interface{}, 0, len(attrs))
-	for _, attr := range attrs {
-		ifattrs = append(ifattrs, attr)
-	}
 	opts := []tracer.SpanOption{
-		tracer.WithSpanLabels(ifattrs...),
+		tracer.WithSpanLabels(attrs...),
 		tracer.WithSpanKind(tracer.SpanKindConsumer),
 	}
 
 	if r.Context == nil {
 		r.Context = context.Background()
 	}
+	md, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		md = metadata.New(len(r.Headers))
+	}
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
 	// Extract the span context from the record.
 	// ctx := t.propagators.Extract(r.Context, NewRecordCarrier(r))
 	// Start the "receive" span.
-	newCtx, _ := m.tracer.Start(r.Context, r.Topic+" receive", opts...)
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewIncomingContext(r.Context, md), r.Topic+" receive", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, r.Topic+" receive", opts...)
+	}
+
 	// Update the record context.
-	r.Context = newCtx
 }
 
 // OnFetchRecordUnbuffered continues and ends the "receive" span for an
 // unbuffered record.
 func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
-	if span, ok := tracer.SpanFromContext(r.Context); ok {
-		defer span.Finish()
-	}
+	span, _ := tracer.SpanFromContext(r.Context)
+	span.Finish()
 }
 
 // WithProcessSpan starts a new span for the "process" operation on a consumer
@@ -168,7 +154,7 @@ func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
 // iteration of your processing for the record.
 func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Span) {
 	// Set up the span options.
-	attrs := []attribute.KeyValue{
+	attrs := []interface{}{
 		semconv.MessagingSystemKey.String("kafka"),
 		semconv.MessagingSourceKindTopic,
 		semconv.MessagingSourceName(r.Topic),
@@ -177,10 +163,6 @@ func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Spa
 		semconv.MessagingKafkaMessageOffset(int(r.Offset)),
 	}
 	attrs = maybeKeyAttr(attrs, r)
-	ifattrs := make([]interface{}, 0, len(attrs))
-	for _, attr := range attrs {
-		ifattrs = append(ifattrs, attr)
-	}
 	if m.clientID != "" {
 		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
 	}
@@ -188,18 +170,26 @@ func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Spa
 		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
 	}
 	opts := []tracer.SpanOption{
-		tracer.WithSpanLabels(ifattrs...),
+		tracer.WithSpanLabels(attrs...),
 		tracer.WithSpanKind(tracer.SpanKindConsumer),
 	}
 
 	if r.Context == nil {
 		r.Context = context.Background()
 	}
+	md, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		md = metadata.New(len(r.Headers))
+	}
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
 	// Start a new span using the provided context and options.
 	return m.tracer.Start(r.Context, r.Topic+" process", opts...)
 }
 
-func maybeKeyAttr(attrs []attribute.KeyValue, r *kgo.Record) []attribute.KeyValue {
+func maybeKeyAttr(attrs []interface{}, r *kgo.Record) []interface{} {
 	if r.Key == nil {
 		return attrs
 	}