update for latest micro

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

.gitignore

@@ -13,3 +13,9 @@
 
 # Dependency directories (remove the comment below to include it)
 # vendor/
+
+# General
+.DS_Store
+.idea
+.vscode
+bin/

README.md

@@ -1,9 +1,2 @@
-# micro-broker-kgo
+# broker-kgo
-yet another micro kafka broker alternative
 
-TODO:
-* dont always append options from context on Init and New
-* add SubscriberOptions(...kgo.Opt)
-* add ServerSubscribeOptions(...kgo.Opt)
-* check PublisherOptions(...kgo.Opt)
-* check ClientPublisherOptions(...kgo.Opt)

carrier.go

@@ -0,0 +1,76 @@
+package kgo
+
+import (
+	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v3/metadata"
+)
+
+// RecordCarrier injects and extracts traces from a kgo.Record.
+//
+// This type exists to satisfy the otel/propagation.TextMapCarrier interface.
+type RecordCarrier struct {
+	record *kgo.Record
+}
+
+// NewRecordCarrier creates a new RecordCarrier.
+func NewRecordCarrier(record *kgo.Record) RecordCarrier {
+	return RecordCarrier{record: record}
+}
+
+// Get retrieves a single value for a given key if it exists.
+func (c RecordCarrier) Get(key string) string {
+	for _, h := range c.record.Headers {
+		if h.Key == key {
+			return string(h.Value)
+		}
+	}
+	return ""
+}
+
+// Set sets a header.
+func (c RecordCarrier) Set(key, val string) {
+	// Check if key already exists.
+	for i, h := range c.record.Headers {
+		if h.Key == key {
+			// Key exist, update the value.
+			c.record.Headers[i].Value = []byte(val)
+			return
+		}
+	}
+	// Key does not exist, append new header.
+	c.record.Headers = append(c.record.Headers, kgo.RecordHeader{
+		Key:   key,
+		Value: []byte(val),
+	})
+}
+
+// Keys returns a slice of all key identifiers in the carrier.
+func (c RecordCarrier) Keys() []string {
+	out := make([]string, len(c.record.Headers))
+	for i, h := range c.record.Headers {
+		out[i] = h.Key
+	}
+	return out
+}
+
+func setHeaders(r *kgo.Record, md metadata.Metadata) {
+	seen := make(map[string]struct{})
+loop:
+	for k, v := range md {
+		for i := 0; i < len(r.Headers); i++ {
+			if r.Headers[i].Key == k {
+				// Key exist, update the value.
+				r.Headers[i].Value = []byte(v)
+				continue loop
+			} else if _, ok := seen[k]; ok {
+				continue loop
+			}
+			// Key does not exist, append new header.
+			r.Headers = append(r.Headers, kgo.RecordHeader{
+				Key:   k,
+				Value: []byte(v),
+			})
+			seen[k] = struct{}{}
+		}
+	}
+}

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

@@ -1,14 +1,25 @@
 module go.unistack.org/micro-broker-kgo/v3
 
-go 1.17
+go 1.22.7
+
+toolchain go1.23.3
 
 require (
-	github.com/twmb/franz-go v1.11.5
+	github.com/google/uuid v1.6.0
-	github.com/twmb/franz-go/pkg/kmsg v1.3.0
+	github.com/twmb/franz-go v1.18.0
-	go.unistack.org/micro/v3 v3.10.14
+	github.com/twmb/franz-go/pkg/kadm v1.14.0
+	github.com/twmb/franz-go/pkg/kmsg v1.9.0
+	go.opentelemetry.io/otel v1.32.0
+	go.unistack.org/micro/v3 v3.11.0
 )
 
 require (
-	github.com/klauspost/compress v1.15.9 // indirect
+	github.com/klauspost/compress v1.17.11 // indirect
-	github.com/pierrec/lz4/v4 v4.1.15 // indirect
+	github.com/pierrec/lz4/v4 v4.1.21 // indirect
+	go.unistack.org/micro-proto/v3 v3.4.1 // indirect
+	golang.org/x/crypto v0.29.0 // indirect
+	golang.org/x/sys v0.27.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a // indirect
+	google.golang.org/grpc v1.68.0 // indirect
+	google.golang.org/protobuf v1.35.2 // indirect
 )

go.sum

@@ -1,24 +1,28 @@
-github.com/imdario/mergo v0.3.13/go.mod h1:4lJ1jqUDcsbIECGy0RUJAXNIhg+6ocWgb1ALK2O4oXg=
+github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
-github.com/klauspost/compress v1.15.9 h1:wKRjX6JRtDdrE9qwa4b/Cip7ACOshUI4smpCQanqjSY=
+github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
-github.com/klauspost/compress v1.15.9/go.mod h1:PhcZ0MbTNciWF3rruxRgKxI5NkcHHrHUDtV4Yw2GlzU=
+github.com/klauspost/compress v1.17.11 h1:In6xLpyWOi1+C7tXUUWv2ot1QvBjxevKAaI6IXrJmUc=
-github.com/patrickmn/go-cache v2.1.0+incompatible/go.mod h1:3Qf8kWWT7OJRJbdiICTKqZju1ZixQ/KpMGzzAfe6+WQ=
+github.com/klauspost/compress v1.17.11/go.mod h1:pMDklpSncoRMuLFrf1W9Ss9KT+0rH90U12bZKk7uwG0=
-github.com/pierrec/lz4/v4 v4.1.15 h1:MO0/ucJhngq7299dKLwIMtgTfbkoSPF6AoMYDd8Q4q0=
+github.com/pierrec/lz4/v4 v4.1.21 h1:yOVMLb6qSIDP67pl/5F7RepeKYu/VmTyEXvuMI5d9mQ=
-github.com/pierrec/lz4/v4 v4.1.15/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
+github.com/pierrec/lz4/v4 v4.1.21/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
-github.com/silas/dag v0.0.0-20211117232152-9d50aa809f35/go.mod h1:7RTUFBdIRC9nZ7/3RyRNH1bdqIShrDejd1YbLwgPS+I=
+github.com/twmb/franz-go v1.18.0 h1:25FjMZfdozBywVX+5xrWC2W+W76i0xykKjTdEeD2ejw=
-github.com/twmb/franz-go v1.11.5 h1:TTv5lVJd+87XkmP9dWN9Jgpf7IUUr7a7jee+byR8LBE=
+github.com/twmb/franz-go v1.18.0/go.mod h1:zXCGy74M0p5FbXsLeASdyvfLFsBvTubVqctIaa5wQ+I=
-github.com/twmb/franz-go v1.11.5/go.mod h1:FvaHNlpT6woVYIl6LAuIeL7yHol1Fp6Gv2Dn21AvH78=
+github.com/twmb/franz-go/pkg/kadm v1.14.0 h1:nAn1co1lXzJQocpzyIyOFOjUBf4WHWs5/fTprXy2IZs=
-github.com/twmb/franz-go/pkg/kmsg v1.3.0 h1:ouBETB7nTqRxiO5E8/pySoFZtVEW2VWw55z3/bsUzTw=
+github.com/twmb/franz-go/pkg/kadm v1.14.0/go.mod h1:XjOPz6ZaXXjrW2jVCfLuucP8H1w2TvD6y3PT2M+aAM4=
-github.com/twmb/franz-go/pkg/kmsg v1.3.0/go.mod h1:SxG/xJKhgPu25SamAq0rrucfp7lbzCpEXOC+vH/ELrY=
+github.com/twmb/franz-go/pkg/kmsg v1.9.0 h1:JojYUph2TKAau6SBtErXpXGC7E3gg4vGZMv9xFU/B6M=
-go.unistack.org/micro/v3 v3.10.14 h1:7fgLpwGlCN67twhwtngJDEQvrMkUBDSA5vzZqxIDqNE=
+github.com/twmb/franz-go/pkg/kmsg v1.9.0/go.mod h1:CMbfazviCyY6HM0SXuG5t9vOwYDHRCSrJJyBAe5paqg=
-go.unistack.org/micro/v3 v3.10.14/go.mod h1:uMAc0U/x7dmtICCrblGf0ZLgYegu3VwQAquu+OFCw1Q=
+go.opentelemetry.io/otel v1.32.0 h1:WnBN+Xjcteh0zdk01SVqV55d/m62NJLJdIyb4y/WO5U=
-golang.org/x/crypto v0.0.0-20220817201139-bc19a97f63c8/go.mod h1:IxCIyHEi3zRg3s0A5j5BB6A9Jmi73HwBIUl50j+osU4=
+go.opentelemetry.io/otel v1.32.0/go.mod h1:00DCVSB0RQcnzlwyTfqtxSm+DRr9hpYrHjNGiBHVQIg=
-golang.org/x/net v0.0.0-20211112202133-69e39bad7dc2/go.mod h1:9nx3DQGgdP8bBQD5qxJ1jj9UTztislL4KSBs9R2vV5Y=
+go.unistack.org/micro-proto/v3 v3.4.1 h1:UTjLSRz2YZuaHk9iSlVqqsA50JQNAEK2ZFboGqtEa9Q=
-golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+go.unistack.org/micro-proto/v3 v3.4.1/go.mod h1:okx/cnOhzuCX0ggl/vToatbCupi0O44diiiLLsZ93Zo=
-golang.org/x/sys v0.0.0-20210423082822-04245dca01da/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+go.unistack.org/micro/v3 v3.11.0 h1:usQ+8wQuOWpQd4+DGhFXSgZ+e+wOBjuT3W5GJZ02bSs=
-golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+go.unistack.org/micro/v3 v3.11.0/go.mod h1:YzMldzHN9Ei+zy5t/Psu7RUWDZwUfrNYiStSQtTz90g=
-golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
+golang.org/x/crypto v0.29.0 h1:L5SG1JTTXupVV3n6sUqMTeWbjAyfPwoda2DLX8J8FrQ=
-golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
+golang.org/x/crypto v0.29.0/go.mod h1:+F4F4N5hv6v38hfeYwTdx20oUvLLc+QfrE9Ax9HtgRg=
-golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
+golang.org/x/sys v0.27.0 h1:wBqf8DvsY9Y/2P8gAfPDEYNuS30J4lPHJxXSb/nJZ+s=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
+golang.org/x/sys v0.27.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
-gopkg.in/yaml.v3 v3.0.0/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a h1:hgh8P4EuoxpsuKMXX/To36nOFD7vixReXgn8lPGnt+o=
-gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a/go.mod h1:5uTbfoYQed2U9p3KIj2/Zzm02PYhndfdmML0qC3q3FU=
+google.golang.org/grpc v1.68.0 h1:aHQeeJbo8zAkAa3pRzrVjZlbz6uSfeOXlJNQM0RAbz0=
+google.golang.org/grpc v1.68.0/go.mod h1:fmSPC5AsjSBCK54MyHRx48kpOti1/jRfOlwEWywNjWA=
+google.golang.org/protobuf v1.35.2 h1:8Ar7bF+apOIoThw1EdZl0p1oWvMqTHmpA2fRTyZO8io=
+google.golang.org/protobuf v1.35.2/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=

kadmtest.go

@@ -0,0 +1,65 @@
+//go:build ignore
+
+package main
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"time"
+
+	"github.com/twmb/franz-go/pkg/kadm"
+	"github.com/twmb/franz-go/pkg/kgo"
+	"github.com/twmb/franz-go/pkg/kversion"
+
+	//"github.com/twmb/franz-go/pkg/sasl/scram"
+	"github.com/twmb/franz-go/pkg/sasl/plain"
+)
+
+func die(msg string, args ...any) {
+	fmt.Fprintf(os.Stderr, msg, args...)
+	os.Exit(1)
+}
+
+func main() {
+	seeds := []string{"vm-kafka-ump01tn.mbrd.ru:9092", "vm-kafka-ump02tn.mbrd.ru:9092", "vm-kafka-ump03tn.mbrd.ru:9092"}
+
+	pass := "XXXXX"
+	user := "XXXXX"
+
+	var adminClient *kadm.Client
+	{
+		client, err := kgo.NewClient(
+			kgo.SeedBrokers(seeds...),
+			// kgo.SASL((scram.Auth{User: user, Pass: pass}).AsSha512Mechanism()),
+			kgo.SASL((plain.Auth{User: user, Pass: pass}).AsMechanism()),
+
+			// Do not try to send requests newer than 2.4.0 to avoid breaking changes in the request struct.
+			// Sometimes there are breaking changes for newer versions where more properties are required to set.
+			kgo.MaxVersions(kversion.V2_4_0()),
+		)
+		if err != nil {
+			panic(err)
+		}
+		defer client.Close()
+
+		adminClient = kadm.NewClient(client)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
+	defer cancel()
+
+	dg, err := adminClient.DescribeGroups(ctx, "interestrate_loader")
+	if err != nil {
+		die("failed to describe group: %v", err)
+	}
+
+	for _, m := range dg["interestrate_loader"].Members {
+		mc, _ := m.Assigned.AsConsumer()
+		for _, mt := range mc.Topics {
+			for _, p := range mt.Partitions {
+				fmt.Printf("client:%s\tpartitions: %d\n", m.ClientID, p)
+			}
+		}
+	}
+}

kgo.go

@@ -1,20 +1,25 @@
 // Package kgo provides a kafka broker using kgo
-package kgo // import "go.unistack.org/micro-broker-kgo/v3"
+package kgo
 
 import (
 	"context"
 	"errors"
 	"fmt"
 	"math/rand"
+	"net/http"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
+	"github.com/google/uuid"
 	"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"
-	id "go.unistack.org/micro/v3/util/id"
+	"go.unistack.org/micro/v3/semconv"
+	"go.unistack.org/micro/v3/tracer"
 	mrand "go.unistack.org/micro/v3/util/rand"
 )
 
@@ -53,13 +58,25 @@ var DefaultRetryBackoffFn = func() func(int) time.Duration {
 }()
 
 type Broker struct {
+	init      bool
 	c         *kgo.Client
 	kopts     []kgo.Opt
-	connected bool
+	connected *atomic.Uint32
-	init      bool
 	sync.RWMutex
 	opts broker.Options
-	subs []*subscriber
+	subs []*Subscriber
+}
+
+func (r *Broker) Live() bool {
+	return r.connected.Load() == 1
+}
+
+func (r *Broker) Ready() bool {
+	return r.connected.Load() == 1
+}
+
+func (r *Broker) Health() bool {
+	return r.connected.Load() == 1
 }
 
 func (k *Broker) Address() string {
@@ -70,63 +87,91 @@ 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
 
+	sp, _ := tracer.SpanFromContext(ctx)
+
+	clientID := "kgo"
+	group := ""
+	if k.opts.Context != nil {
+		if id, ok := k.opts.Context.Value(clientIDKey{}).(string); ok {
+			clientID = id
+		}
+		if id, ok := k.opts.Context.Value(groupKey{}).(string); ok {
+			group = id
+		}
+	}
+
+	htracer := &hookTracer{group: group, clientID: clientID, tracer: k.opts.Tracer}
+	opts = append(opts,
+		kgo.WithHooks(&hookMeter{meter: k.opts.Meter}),
+		kgo.WithHooks(htracer),
+	)
+
 	select {
 	case <-ctx.Done():
-		return nil, ctx.Err()
+		if ctx.Err() != nil {
+			if sp != nil {
+				sp.SetStatus(tracer.SpanStatusError, ctx.Err().Error())
+			}
+		}
+		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 {
-			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 {
-	k.RLock()
+	if k.connected.Load() == 1 {
-	if k.connected {
-		k.RUnlock()
 		return nil
 	}
-	k.RUnlock()
 
 	nctx := k.opts.Context
 	if ctx != nil {
 		nctx = ctx
 	}
 
-	c, err := k.connect(nctx, k.kopts...)
+	c, _, err := k.connect(nctx, k.kopts...)
 	if err != nil {
 		return err
 	}
 
 	k.Lock()
 	k.c = c
-	k.connected = true
+	k.connected.Store(1)
 	k.Unlock()
 
 	return nil
 }
 
 func (k *Broker) Disconnect(ctx context.Context) error {
-	k.RLock()
+	if k.connected.Load() == 0 {
-	if !k.connected {
-		k.RUnlock()
 		return nil
 	}
-	k.RUnlock()
 
 	nctx := k.opts.Context
 	if ctx != nil {
 		nctx = ctx
 	}
+	var span tracer.Span
+	ctx, span = k.opts.Tracer.Start(ctx, "Disconnect")
+	defer span.Finish()
 
 	k.Lock()
 	defer k.Unlock()
@@ -148,7 +193,7 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 		}
 	}
 
-	k.connected = false
+	k.connected.Store(0)
 	return nil
 }
 
@@ -156,6 +201,10 @@ func (k *Broker) Init(opts ...broker.Option) error {
 	k.Lock()
 	defer k.Unlock()
 
+	if len(opts) == 0 && k.init {
+		return nil
+	}
+
 	for _, o := range opts {
 		o(&k.opts)
 	}
@@ -179,6 +228,8 @@ func (k *Broker) Init(opts ...broker.Option) error {
 		}
 	}
 
+	k.init = true
+
 	return nil
 }
 
@@ -196,42 +247,43 @@ 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 {
-	k.RLock()
+	k.Lock()
-	if !k.connected {
+	if k.connected.Load() == 0 {
-		k.RUnlock()
+		c, _, err := k.connect(ctx, k.kopts...)
-		k.Lock()
-		c, err := k.connect(ctx, k.kopts...)
 		if err != nil {
 			k.Unlock()
 			return err
 		}
 		k.c = c
-		k.connected = true
+		k.connected.Store(1)
-		k.Unlock()
 	}
-	k.RUnlock()
+	k.Unlock()
 
 	options := broker.NewPublishOptions(opts...)
 	records := make([]*kgo.Record, 0, len(msgs))
 	var errs []string
 	var err error
 	var key []byte
+	var promise func(*kgo.Record, error)
 
 	if options.Context != nil {
 		if k, ok := options.Context.Value(publishKey{}).([]byte); ok && k != nil {
 			key = k
 		}
+		if p, ok := options.Context.Value(publishPromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
+			promise = p
+		}
 	}
 
 	for _, msg := range msgs {
 		rec := &kgo.Record{Context: ctx, Key: key}
 		rec.Topic, _ = msg.Header.Get(metadata.HeaderTopic)
-		if options.BodyOnly {
+		msg.Header.Del(metadata.HeaderTopic)
-			rec.Value = msg.Body
+		k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Inc()
-		} else if k.opts.Codec.String() == "noop" {
+		if options.BodyOnly || k.opts.Codec.String() == "noop" {
 			rec.Value = msg.Body
 			for k, v := range msg.Header {
-				rec.Headers = append(rec.Headers, kgo.RecordHeader{Key: k, Value: []byte(v)})
+				rec.Headers = append(rec.Headers, kgo.RecordHeader{Key: http.CanonicalHeaderKey(k), Value: []byte(v)})
 			}
 		} else {
 			rec.Value, err = k.opts.Codec.Marshal(msg)
@@ -242,10 +294,36 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 		records = append(records, rec)
 	}
 
+	if promise != nil {
+		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(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(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "failure").Inc()
+				} else {
+					k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "success").Inc()
+				}
+				promise(r, err)
+			})
+		}
+		return nil
+	}
+	ts := time.Now()
 	results := k.c.ProduceSync(ctx, records...)
+	te := time.Since(ts)
 	for _, result := range results {
+		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(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "failure").Inc()
 			errs = append(errs, result.Err.Error())
+		} else {
+			k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "success").Inc()
 		}
 	}
 
@@ -256,6 +334,22 @@ 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
 }
@@ -264,11 +358,11 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	options := broker.NewSubscribeOptions(opts...)
 
 	if options.Group == "" {
-		uid, err := id.New()
+		uid, err := uuid.NewRandom()
 		if err != nil {
 			return nil, err
 		}
-		options.Group = uid
+		options.Group = uid.String()
 	}
 
 	commitInterval := DefaultCommitInterval
@@ -278,7 +372,7 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	sub := &subscriber{
+	sub := &Subscriber{
 		topic:     topic,
 		opts:      options,
 		handler:   handler,
@@ -305,7 +399,7 @@ 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
 	}
@@ -323,6 +417,8 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	}
 
 	sub.c = c
+	sub.htracer = htracer
+
 	go sub.poll(ctx)
 
 	k.Lock()
@@ -336,7 +432,6 @@ func (k *Broker) String() string {
 }
 
 func NewBroker(opts ...broker.Option) *Broker {
-	rand.Seed(time.Now().Unix())
 	options := broker.NewOptions(opts...)
 
 	kaddrs := options.Addrs
@@ -349,14 +444,13 @@ 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.WithAddCallerSkipCount(2)), ctx: options.Context}),
-		// kgo.WithLogger(kgo.BasicLogger(os.Stderr, kgo.LogLevelDebug, func() string { return time.Now().Format(time.StampMilli) })),
-		kgo.WithHooks(&metrics{meter: options.Meter}),
 		kgo.SeedBrokers(kaddrs...),
 		kgo.RetryBackoffFn(DefaultRetryBackoffFn),
 		kgo.BlockRebalanceOnPoll(),
 		kgo.Balancers(kgo.CooperativeStickyBalancer()),
 		kgo.FetchIsolationLevel(kgo.ReadUncommitted()),
+		kgo.UnknownTopicRetries(0),
 	}
 
 	if options.Context != nil {

kgo_test.go

@@ -50,7 +50,7 @@ func TestPubSub(t *testing.T) {
 		t.Skip()
 	}
 
-	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel), logger.WithCallerSkipCount(3)); err != nil {
+	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel)); err != nil {
 		t.Fatal(err)
 	}
 	ctx := context.Background()

logger.go

@@ -2,7 +2,6 @@ package kgo
 
 import (
 	"context"
-	"fmt"
 
 	"github.com/twmb/franz-go/pkg/kgo"
 	"go.unistack.org/micro/v3/logger"
@@ -29,15 +28,8 @@ func (l *mlogger) Log(lvl kgo.LogLevel, msg string, args ...interface{}) {
 	default:
 		return
 	}
-	if len(args) > 0 {
+
-		fields := make(map[string]interface{}, int(len(args)/2))
+	l.l.Log(l.ctx, mlvl, msg, args...)
-		for i := 0; i <= len(args)/2; i += 2 {
-			fields[fmt.Sprintf("%v", args[i])] = args[i+1]
-		}
-		l.l.Fields(fields).Log(l.ctx, mlvl, msg)
-	} else {
-		l.l.Log(l.ctx, mlvl, msg)
-	}
 }
 
 func (l *mlogger) Level() kgo.LogLevel {

meter.go

@@ -9,19 +9,26 @@ import (
 	"go.unistack.org/micro/v3/meter"
 )
 
-type metrics struct {
+type hookMeter struct {
 	meter meter.Meter
 }
 
 var (
-	_ kgo.HookBrokerConnect       = &metrics{}
+	_ kgo.HookBrokerConnect    = &hookMeter{}
-	_ kgo.HookBrokerDisconnect    = &metrics{}
+	_ kgo.HookBrokerDisconnect = &hookMeter{}
-	_ kgo.HookBrokerRead          = &metrics{}
+	// HookBrokerE2E
-	_ kgo.HookBrokerThrottle      = &metrics{}
+	_ kgo.HookBrokerRead     = &hookMeter{}
-	_ kgo.HookBrokerWrite         = &metrics{}
+	_ kgo.HookBrokerThrottle = &hookMeter{}
-	_ kgo.HookFetchBatchRead      = &metrics{}
+	_ kgo.HookBrokerWrite    = &hookMeter{}
-	_ kgo.HookProduceBatchWritten = &metrics{}
+	_ kgo.HookFetchBatchRead = &hookMeter{}
-	_ kgo.HookGroupManageError    = &metrics{}
+	// HookFetchRecordBuffered
+	// HookFetchRecordUnbuffered
+	_ kgo.HookGroupManageError = &hookMeter{}
+	// HookNewClient
+	_ kgo.HookProduceBatchWritten = &hookMeter{}
+	// HookProduceRecordBuffered
+	// HookProduceRecordPartitioned
+	// HookProduceRecordUnbuffered
 )
 
 const (
@@ -54,11 +61,11 @@ const (
 	labelTopic   = "topic"
 )
 
-func (m *metrics) OnGroupManageError(err error) {
+func (m *hookMeter) OnGroupManageError(err error) {
 	m.meter.Counter(metricBrokerGroupErrors).Inc()
 }
 
-func (m *metrics) OnBrokerConnect(meta kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
+func (m *hookMeter) OnBrokerConnect(meta kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
 	node := strconv.Itoa(int(meta.NodeID))
 	if err != nil {
 		m.meter.Counter(metricBrokerConnects, labelNode, node, labelStatus, labelFaulure).Inc()
@@ -67,12 +74,12 @@ func (m *metrics) OnBrokerConnect(meta kgo.BrokerMetadata, _ time.Duration, _ ne
 	m.meter.Counter(metricBrokerConnects, labelNode, node, labelStatus, labelSuccess).Inc()
 }
 
-func (m *metrics) OnBrokerDisconnect(meta kgo.BrokerMetadata, _ net.Conn) {
+func (m *hookMeter) OnBrokerDisconnect(meta kgo.BrokerMetadata, _ net.Conn) {
 	node := strconv.Itoa(int(meta.NodeID))
 	m.meter.Counter(metricBrokerDisconnects, labelNode, node).Inc()
 }
 
-func (m *metrics) OnBrokerWrite(meta kgo.BrokerMetadata, _ int16, bytesWritten int, writeWait, timeToWrite time.Duration, err error) {
+func (m *hookMeter) OnBrokerWrite(meta kgo.BrokerMetadata, _ int16, bytesWritten int, writeWait, timeToWrite time.Duration, err error) {
 	node := strconv.Itoa(int(meta.NodeID))
 	if err != nil {
 		m.meter.Counter(metricBrokerWriteErrors, labelNode, node).Inc()
@@ -83,7 +90,7 @@ func (m *metrics) OnBrokerWrite(meta kgo.BrokerMetadata, _ int16, bytesWritten i
 	m.meter.Histogram(metricBrokerWriteLatencies, labelNode, node).Update(timeToWrite.Seconds())
 }
 
-func (m *metrics) OnBrokerRead(meta kgo.BrokerMetadata, _ int16, bytesRead int, readWait, timeToRead time.Duration, err error) {
+func (m *hookMeter) OnBrokerRead(meta kgo.BrokerMetadata, _ int16, bytesRead int, readWait, timeToRead time.Duration, err error) {
 	node := strconv.Itoa(int(meta.NodeID))
 	if err != nil {
 		m.meter.Counter(metricBrokerReadErrors, labelNode, node).Inc()
@@ -95,18 +102,18 @@ func (m *metrics) OnBrokerRead(meta kgo.BrokerMetadata, _ int16, bytesRead int,
 	m.meter.Histogram(metricBrokerReadLatencies, labelNode, node).Update(timeToRead.Seconds())
 }
 
-func (m *metrics) OnBrokerThrottle(meta kgo.BrokerMetadata, throttleInterval time.Duration, _ bool) {
+func (m *hookMeter) OnBrokerThrottle(meta kgo.BrokerMetadata, throttleInterval time.Duration, _ bool) {
 	node := strconv.Itoa(int(meta.NodeID))
 	m.meter.Histogram(metricBrokerThrottleLatencies, labelNode, node).Update(throttleInterval.Seconds())
 }
 
-func (m *metrics) OnProduceBatchWritten(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.ProduceBatchMetrics) {
+func (m *hookMeter) OnProduceBatchWritten(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.ProduceBatchMetrics) {
 	node := strconv.Itoa(int(meta.NodeID))
 	m.meter.Counter(metricBrokerProduceBytesUncompressed, labelNode, node, labelTopic, topic).Add(kmetrics.UncompressedBytes)
 	m.meter.Counter(metricBrokerProduceBytesCompressed, labelNode, node, labelTopic, topic).Add(kmetrics.CompressedBytes)
 }
 
-func (m *metrics) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.FetchBatchMetrics) {
+func (m *hookMeter) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.FetchBatchMetrics) {
 	node := strconv.Itoa(int(meta.NodeID))
 	m.meter.Counter(metricBrokerFetchBytesUncompressed, labelNode, node, labelTopic, topic).Add(kmetrics.UncompressedBytes)
 	m.meter.Counter(metricBrokerFetchBytesCompressed, labelNode, node, labelTopic, topic).Add(kmetrics.CompressedBytes)

options.go

@@ -9,8 +9,17 @@ import (
 	"go.unistack.org/micro/v3/client"
 )
 
-// DefaultCommitInterval specifies how fast send commit offsets to kafka
+var (
-var DefaultCommitInterval = 5 * time.Second
+
+	// 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{}
 
@@ -63,6 +72,18 @@ func SubscribeOptions(opts ...kgo.Opt) broker.SubscribeOption {
 	}
 }
 
+type clientIDKey struct{}
+
+func ClientID(id string) broker.Option {
+	return broker.SetOption(clientIDKey{}, id)
+}
+
+type groupKey struct{}
+
+func Group(id string) broker.Option {
+	return broker.SetOption(groupKey{}, id)
+}
+
 type commitIntervalKey struct{}
 
 // CommitInterval specifies interval to send commits
@@ -70,11 +91,21 @@ func CommitInterval(td time.Duration) broker.Option {
 	return broker.SetOption(commitIntervalKey{}, td)
 }
 
-var DefaultSubscribeMaxInflight = 10
-
 type subscribeMaxInflightKey struct{}
 
 // SubscribeMaxInFlight max queued messages
 func SubscribeMaxInFlight(n int) broker.SubscribeOption {
 	return broker.SetSubscribeOption(subscribeMaxInflightKey{}, n)
 }
+
+type publishPromiseKey struct{}
+
+// PublishPromise set the kafka promise func for Produce
+func PublishPromise(fn func(*kgo.Record, error)) broker.PublishOption {
+	return broker.SetPublishOption(publishPromiseKey{}, fn)
+}
+
+// ClientPublishKey set the kafka message key (client option)
+func ClientPublishPromise(fn func(*kgo.Record, error)) client.PublishOption {
+	return client.SetPublishOption(publishPromiseKey{}, fn)
+}

subscriber.go

@@ -2,12 +2,18 @@ package kgo
 
 import (
 	"context"
+	"fmt"
+	"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 +25,7 @@ type consumer struct {
 	c         *kgo.Client
 	topic     string
 	partition int32
+	htracer   *hookTracer
 	opts      broker.SubscribeOptions
 	kopts     broker.Options
 	handler   broker.Handler
@@ -27,9 +34,10 @@ 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
@@ -39,23 +47,28 @@ 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
 	}
 	select {
-	case <-ctx.Done():
+	// case <-ctx.Done():
-		return ctx.Err()
+	//	return ctx.Err()
 	default:
 		s.c.PauseFetchTopics(s.topic)
+		s.c.CloseAllowingRebalance()
 		kc := make(map[string][]int32)
 		for ctp := range s.consumers {
 			kc[ctp.t] = append(kc[ctp.t], ctp.p)
@@ -63,24 +76,59 @@ func (s *subscriber) Unsubscribe(ctx context.Context) error {
 		s.killConsumers(ctx, kc)
 		close(s.done)
 		s.closed = true
+		s.c.ResumeFetchTopics(s.topic)
 	}
 	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():
 			s.c.CloseAllowingRebalance()
 			return
 		case <-s.done:
-			s.c.CloseAllowingRebalance()
 			return
 		default:
 			fetches := s.c.PollRecords(ctx, maxInflight)
@@ -89,7 +137,7 @@ func (s *subscriber) poll(ctx context.Context) {
 				return
 			}
 			fetches.EachError(func(t string, p int32, err error) {
-				s.kopts.Logger.Fatalf(ctx, "[kgo] fetch topic %s partition %d err: %v", t, p, err)
+				s.kopts.Logger.Fatal(ctx, fmt.Sprintf("[kgo] fetch topic %s partition %d error", t, p), err)
 			})
 
 			fetches.EachPartition(func(p kgo.FetchTopicPartition) {
@@ -101,7 +149,7 @@ 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()
 
@@ -111,42 +159,50 @@ func (s *subscriber) killConsumers(ctx context.Context, lost map[string][]int32)
 			pc := s.consumers[tp]
 			delete(s.consumers, tp)
 			close(pc.quit)
-			s.kopts.Logger.Debugf(ctx, "[kgo] waiting for work to finish topic %s partition %d", topic, partition)
+			if s.kopts.Logger.V(logger.DebugLevel) {
+				s.kopts.Logger.Debug(ctx, fmt.Sprintf("[kgo] waiting for work to finish topic %s partition %d", topic, partition))
+			}
 			wg.Add(1)
 			go func() { <-pc.done; wg.Done() }()
 		}
 	}
 }
 
-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)
+	if s.kopts.Logger.V(logger.DebugLevel) {
+		s.kopts.Logger.Debug(ctx, fmt.Sprintf("[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)
+	if s.kopts.Logger.V(logger.DebugLevel) {
+		s.kopts.Logger.Debug(ctx, fmt.Sprintf("[kgo] revoked %#+v", revoked))
+	}
 	s.killConsumers(ctx, revoked)
 	if err := c.CommitMarkedOffsets(ctx); err != nil {
-		s.kopts.Logger.Errorf(ctx, "[kgo] revoked CommitMarkedOffsets err: %v", err)
+		s.kopts.Logger.Error(ctx, "[kgo] revoked CommitMarkedOffsets error", err)
 	}
 }
 
-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,
-
+				htracer:   s.htracer,
-				quit:    make(chan struct{}),
+				quit:      make(chan struct{}),
-				done:    make(chan struct{}),
+				done:      make(chan struct{}),
-				recs:    make(chan kgo.FetchTopicPartition, 4),
+				recs:      make(chan kgo.FetchTopicPartition, 100),
-				handler: s.handler,
+				handler:   s.handler,
-				kopts:   s.kopts,
+				kopts:     s.kopts,
-				opts:    s.opts,
+				opts:      s.opts,
 			}
+			s.Lock()
 			s.consumers[tp{topic, partition}] = pc
+			s.Unlock()
 			go pc.consume()
 		}
 	}
@@ -154,8 +210,10 @@ func (s *subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[str
 
 func (pc *consumer) consume() {
 	defer close(pc.done)
-	pc.kopts.Logger.Debugf(pc.kopts.Context, "starting, topic %s partition %d", pc.topic, pc.partition)
+	if pc.kopts.Logger.V(logger.DebugLevel) {
-	defer pc.kopts.Logger.Debugf(pc.kopts.Context, "killing, topic %s partition %d", pc.topic, pc.partition)
+		pc.kopts.Logger.Debug(pc.kopts.Context, fmt.Sprintf("starting, topic %s partition %d", pc.topic, pc.partition))
+		defer pc.kopts.Logger.Debug(pc.kopts.Context, fmt.Sprintf("killing, topic %s partition %d", pc.topic, pc.partition))
+	}
 
 	eh := pc.kopts.ErrorHandler
 	if pc.opts.ErrorHandler != nil {
@@ -168,66 +226,99 @@ func (pc *consumer) consume() {
 			return
 		case p := <-pc.recs:
 			for _, record := range p.Records {
+				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
 				p.topic = record.Topic
 				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))
+				}
 				if pc.kopts.Codec.String() == "noop" {
-					p.msg.Header = metadata.New(len(record.Headers))
-					for _, hdr := range record.Headers {
-						p.msg.Header.Set(hdr.Key, string(hdr.Value))
-					}
 					p.msg.Body = record.Value
 				} else if pc.opts.BodyOnly {
 					p.msg.Body = record.Value
 				} else {
-					if err := pc.kopts.Codec.Unmarshal(record.Value, p.msg); err != nil {
+					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(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 							if p.ack {
 								pc.c.MarkCommitRecords(record)
 							} else {
 								eventPool.Put(p)
-								pc.kopts.Logger.Infof(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+								pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
 								return
 							}
 							eventPool.Put(p)
+							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.Error(pc.kopts.Context, "[kgo]: unmarshal error", err)
-								pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: failed to unmarshal: %v", err)
-							}
 						}
+						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.Infof(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
+						pc.kopts.Logger.Fatal(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(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "success").Inc()
+				} else {
+					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(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)
+							pc.kopts.Logger.Error(pc.kopts.Context, "[kgo]: subscriber error", err)
 						}
 					}
 				}
+				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?")
+					pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+					sp.SetStatus(tracer.SpanStatusError, "ErrLostMessage")
+					sp.Finish()
 					return
 				}
+				sp.Finish()
 			}
 		}
 	}

tracer.go

@@ -0,0 +1,204 @@
+package kgo
+
+import (
+	"context"
+	"unicode/utf8"
+
+	"github.com/twmb/franz-go/pkg/kgo"
+	semconv "go.opentelemetry.io/otel/semconv/v1.18.0"
+	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/tracer"
+)
+
+type hookTracer struct {
+	clientID string
+	group    string
+	tracer   tracer.Tracer
+}
+
+var messagingSystem = semconv.MessagingSystemKey.String("kafka")
+
+var (
+	_ kgo.HookProduceRecordBuffered   = (*hookTracer)(nil)
+	_ kgo.HookProduceRecordUnbuffered = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordBuffered     = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordUnbuffered   = (*hookTracer)(nil)
+)
+
+// OnProduceRecordBuffered starts a new span for the "publish" operation on a
+// buffered record.
+//
+// It sets span options and injects the span context into record and updates
+// the record's context, so it can be ended in the OnProduceRecordUnbuffered
+// hook.
+func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
+	// Set up span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingDestinationKindTopic,
+		semconv.MessagingDestinationName(r.Topic),
+		semconv.MessagingOperationPublish,
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	opts := []tracer.SpanOption{
+		tracer.WithSpanLabels(attrs...),
+		tracer.WithSpanKind(tracer.SpanKindProducer),
+	}
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	omd, ok := metadata.FromOutgoingContext(r.Context)
+	if !ok {
+		omd = metadata.New(len(r.Headers))
+	}
+
+	md := metadata.Copy(omd)
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewOutgoingContext(r.Context, md), "sdk.broker", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
+	}
+
+	setHeaders(r, omd)
+}
+
+// OnProduceRecordUnbuffered continues and ends the "publish" span for an
+// unbuffered 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) {
+	if span, ok := tracer.SpanFromContext(r.Context); ok {
+		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
+// buffered record.
+//
+// It sets the span options and extracts the span context from the record,
+// updates the record's context to ensure it can be ended in the
+// OnFetchRecordUnbuffered hook and can be used in downstream consumer
+// processing.
+func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
+	// Set up the span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingSourceKindTopic,
+		semconv.MessagingSourceName(r.Topic),
+		semconv.MessagingOperationReceive,
+		semconv.MessagingKafkaSourcePartition(int(r.Partition)),
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	if m.group != "" {
+		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
+	}
+	opts := []tracer.SpanOption{
+		tracer.WithSpanLabels(attrs...),
+		tracer.WithSpanKind(tracer.SpanKindConsumer),
+	}
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+	omd, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		omd = metadata.New(len(r.Headers))
+	}
+
+	md := metadata.Copy(omd)
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewIncomingContext(r.Context, md), "sdk.broker", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
+	}
+
+	setHeaders(r, omd)
+}
+
+// OnFetchRecordUnbuffered continues and ends the "receive" span for an
+// unbuffered record.
+func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
+	span, _ := tracer.SpanFromContext(r.Context)
+	span.Finish()
+}
+
+// WithProcessSpan starts a new span for the "process" operation on a consumer
+// record.
+//
+// It sets up the span options. The user's application code is responsible for
+// ending the span.
+//
+// This should only ever be called within a polling loop of a consumed record and
+// not a record which has been created for producing, so call this at the start of each
+// iteration of your processing for the record.
+func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Span) {
+	// Set up the span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingSourceKindTopic,
+		semconv.MessagingSourceName(r.Topic),
+		semconv.MessagingOperationProcess,
+		semconv.MessagingKafkaSourcePartition(int(r.Partition)),
+		semconv.MessagingKafkaMessageOffset(int(r.Offset)),
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	if m.group != "" {
+		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
+	}
+	opts := []tracer.SpanOption{
+		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, "sdk.broker", opts...)
+}
+
+func maybeKeyAttr(attrs []interface{}, r *kgo.Record) []interface{} {
+	if r.Key == nil {
+		return attrs
+	}
+	var keykey string
+	if !utf8.Valid(r.Key) {
+		return attrs
+	}
+	keykey = string(r.Key)
+	return append(attrs, semconv.MessagingKafkaMessageKeyKey.String(keykey))
+}