add ExposeLag option, to be able to disable lag exporting

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

.github/workflows/job_coverage.yml

@@ -8,12 +8,11 @@ on:
       - '.gitea/**'
   pull_request:
     branches: [ main, v3, v4 ]
-  # Allows you to run this workflow manually from the Actions tab
-  workflow_dispatch:
 
 jobs:
 
   build:
+    if: github.server_url != 'https://github.com'
     runs-on: ubuntu-latest
     steps:
     - name: checkout code
@@ -42,8 +41,8 @@ jobs:
       name: autocommit
       with:
         commit_message: Apply Code Coverage Badge
-        skip_fetch: true
+        skip_fetch: false
-        skip_checkout: true
+        skip_checkout: false
         file_pattern: ./README.md
 
     - name: push

.github/workflows/job_lint.yml

@@ -24,6 +24,6 @@ jobs:
     - name: setup deps
       run: go get -v ./...
     - name: run lint
-      uses: https://github.com/golangci/golangci-lint-action@v6
+      uses: golangci/golangci-lint-action@v6
       with:
         version: 'latest'

.github/workflows/job_sync.yml

@@ -8,7 +8,7 @@ on:
 
 jobs:
   sync:
-    if: github.server_url == 'zhttps://github.com'
+    if: github.server_url != 'https://github.com'
     runs-on: ubuntu-latest
     steps:
     - name: init
@@ -18,34 +18,76 @@ jobs:
         echo "machine git.unistack.org login vtolstov password ${{ secrets.TOKEN_GITEA }}" >> /root/.netrc
         echo "machine github.com login vtolstov password ${{ secrets.TOKEN_GITHUB }}" >> /root/.netrc
 
+    - name: check master
+      id: check_master
+      run: |
+        src_hash=$(git ls-remote https://github.com/${GITHUB_REPOSITORY} refs/heads/master | cut -f1)
+        dst_hash=$(git ls-remote ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} refs/heads/master | cut -f1)
+        echo "src_hash=$src_hash"
+        echo "dst_hash=$dst_hash"
+        if [ "$src_hash" != "$dst_hash" ]; then
+          echo "sync_needed=true" >> $GITHUB_OUTPUT
+        else
+          echo "sync_needed=false" >> $GITHUB_OUTPUT
+        fi
+
     - name: sync master
+      if: steps.check_master.outputs.sync_needed == 'true'
       run: |
         git clone --filter=blob:none --filter=tree:0 --branch master --single-branch ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} repo
         cd repo
         git remote add --no-tags --fetch --track master upstream https://github.com/${GITHUB_REPOSITORY}
-        git merge upstream/master
+        git pull --rebase upstream master
         git push upstream master --progress
         git push origin master --progress
         cd ../
         rm -rf repo
 
+    - name: check v3
+      id: check_v3
+      run: |
+        src_hash=$(git ls-remote https://github.com/${GITHUB_REPOSITORY} refs/heads/v3 | cut -f1)
+        dst_hash=$(git ls-remote ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} refs/heads/v3 | cut -f1)
+        echo "src_hash=$src_hash"
+        echo "dst_hash=$dst_hash"
+        if [ "$src_hash" != "$dst_hash" ]; then
+          echo "sync_needed=true" >> $GITHUB_OUTPUT
+        else
+          echo "sync_needed=false" >> $GITHUB_OUTPUT
+        fi
+
     - name: sync v3
+      if: steps.check_v3.outputs.sync_needed == 'true'
       run: |
         git clone --filter=blob:none --filter=tree:0 --branch v3 --single-branch ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} repo
         cd repo
         git remote add --no-tags --fetch --track v3 upstream https://github.com/${GITHUB_REPOSITORY}
-        git merge upstream/v3
+        git pull --rebase upstream v3
         git push upstream v3 --progress
         git push origin v3 --progress
         cd ../
         rm -rf repo
 
+    - name: check v4
+      id: check_v4
+      run: |
+        src_hash=$(git ls-remote https://github.com/${GITHUB_REPOSITORY} refs/heads/v4 | cut -f1)
+        dst_hash=$(git ls-remote ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} refs/heads/v4 | cut -f1)
+        echo "src_hash=$src_hash"
+        echo "dst_hash=$dst_hash"
+        if [ "$src_hash" != "$dst_hash" ]; then
+          echo "sync_needed=true" >> $GITHUB_OUTPUT
+        else
+          echo "sync_needed=false" >> $GITHUB_OUTPUT
+        fi
+
     - name: sync v4
+      if: steps.check_v4.outputs.sync_needed == 'true'
       run: |
         git clone --filter=blob:none --filter=tree:0 --branch v4 --single-branch ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY} repo
         cd repo
         git remote add --no-tags --fetch --track v4 upstream https://github.com/${GITHUB_REPOSITORY}
-        git merge upstream/v4
+        git pull --rebase upstream v4
         git push upstream v4 --progress
         git push origin v4 --progress
         cd ../

.github/workflows/job_tests.yml

@@ -32,19 +32,19 @@ jobs:
         go-version: 'stable'
     - name: setup go work
       env:
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: |
         go work init
         go work use .
         go work use micro-tests
     - name: setup deps
       env:
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: go get -v ./...
     - name: run tests
       env:
         INTEGRATION_TESTS: yes
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: |
         cd micro-tests
         go test -mod readonly -v ./... || true

README.md

@@ -1,9 +1,2 @@
-# broker-kgo
+# micro-broker-kgo
-
+![Coverage](https://img.shields.io/badge/Coverage-66.9%25-yellow)
-TODO:
-
-1) экспортируем текущий оффсет для каждой партиции в группе
-2) экспортируем лаг для группы
-3) мониторим
-   1) если есть лаг больше нуля
-   2) если дельта оффсета за нужное нам время не 

broker.go

@@ -1,69 +0,0 @@
-package kgo
-
-import (
-	"context"
-	"net"
-	"sync/atomic"
-	"time"
-
-	"github.com/twmb/franz-go/pkg/kgo"
-	"go.unistack.org/micro/v4/logger"
-)
-
-type hookEvent struct {
-	log          logger.Logger
-	fatalOnError bool
-	connected    *atomic.Uint32
-}
-
-var (
-	_ kgo.HookBrokerConnect           = &hookEvent{}
-	_ kgo.HookBrokerDisconnect        = &hookEvent{}
-	_ kgo.HookBrokerRead              = &hookEvent{}
-	_ kgo.HookBrokerWrite             = &hookEvent{}
-	_ kgo.HookGroupManageError        = &hookEvent{}
-	_ kgo.HookProduceRecordUnbuffered = &hookEvent{}
-)
-
-func (m *hookEvent) OnGroupManageError(err error) {
-	if err != nil {
-		//		m.connected.Store(0)
-		if m.fatalOnError {
-			m.log.Fatal(context.TODO(), "kgo.OnGroupManageError", err)
-		}
-	}
-}
-
-func (m *hookEvent) OnBrokerConnect(_ kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
-	if err != nil {
-		//		m.connected.Store(0)
-		if m.fatalOnError {
-			m.log.Fatal(context.TODO(), "kgo.OnBrokerConnect", err)
-		}
-	}
-}
-
-func (m *hookEvent) OnBrokerDisconnect(_ kgo.BrokerMetadata, _ net.Conn) {
-	// m.connected.Store(0)
-}
-
-func (m *hookEvent) OnBrokerWrite(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
-	if err != nil {
-		//		m.connected.Store(0)
-		if m.fatalOnError {
-			m.log.Fatal(context.TODO(), "kgo.OnBrokerWrite", err)
-		}
-	}
-}
-
-func (m *hookEvent) OnBrokerRead(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
-	if err != nil {
-		//	m.connected.Store(0)
-	}
-}
-
-func (m *hookEvent) OnProduceRecordUnbuffered(_ *kgo.Record, err error) {
-	if err != nil {
-		//	m.connected.Store(0)
-	}
-}

errors.go

@@ -0,0 +1,10 @@
+package kgo
+
+import (
+	"context"
+	"errors"
+)
+
+func isContextError(err error) bool {
+	return errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded)
+}

go.mod

@@ -1,24 +1,27 @@
 module go.unistack.org/micro-broker-kgo/v4
 
-go 1.23.0
+go 1.24.0
 
 require (
-	github.com/twmb/franz-go v1.18.1
+	github.com/stretchr/testify v1.11.1
-	github.com/twmb/franz-go/pkg/kadm v1.16.0
+	github.com/twmb/franz-go v1.19.5
+	github.com/twmb/franz-go/pkg/kadm v1.16.1
+	github.com/twmb/franz-go/pkg/kfake v0.0.0-20250508175730-72e1646135e3
 	github.com/twmb/franz-go/pkg/kmsg v1.11.2
-	go.opentelemetry.io/otel v1.35.0
+	go.opentelemetry.io/otel v1.38.0
-	go.unistack.org/micro/v4 v4.1.7
+	go.unistack.org/micro/v4 v4.1.21
 )
 
 require (
-	github.com/ash3in/uuidv8 v1.2.0 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/klauspost/compress v1.18.0 // indirect
 	github.com/matoous/go-nanoid v1.5.1 // indirect
 	github.com/pierrec/lz4/v4 v4.1.22 // indirect
-	github.com/spf13/cast v1.7.1 // indirect
+	github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 // indirect
+	github.com/spf13/cast v1.10.0 // indirect
 	go.unistack.org/micro-proto/v4 v4.1.0 // indirect
-	golang.org/x/crypto v0.37.0 // indirect
+	golang.org/x/crypto v0.43.0 // indirect
-	google.golang.org/protobuf v1.36.6 // indirect
+	google.golang.org/protobuf v1.36.10 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -1,7 +1,3 @@
-github.com/DATA-DOG/go-sqlmock v1.5.2 h1:OcvFkGmslmlZibjAjaHm3L//6LiuBgolP7OputlJIzU=
-github.com/DATA-DOG/go-sqlmock v1.5.2/go.mod h1:88MAG/4G7SMwSE3CeA0ZKzrT5CiOU3OJ+JlNzwDqpNU=
-github.com/ash3in/uuidv8 v1.2.0 h1:2oogGdtCPwaVtyvPPGin4TfZLtOGE5F+W++E880G6SI=
-github.com/ash3in/uuidv8 v1.2.0/go.mod h1:BnU0wJBxnzdEKmVg4xckBkD+VZuecTFTUP3M0dWgyY4=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1VwoXQT9A3Wy9MM3WgvqSxFWenqJduM=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/frankban/quicktest v1.14.6 h1:7Xjx+VpznH+oBnejlPUj8oUpdxnVs4f8XU8WnHkI4W8=
@@ -24,26 +20,28 @@ github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 h1:Jamvg5psRI
 github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/rogpeppe/go-internal v1.13.1 h1:KvO1DLK/DRN07sQ1LQKScxyZJuNnedQ5/wKSR38lUII=
 github.com/rogpeppe/go-internal v1.13.1/go.mod h1:uMEvuHeurkdAXX61udpOXGD/AzZDWNMNyH2VO9fmH0o=
-github.com/spf13/cast v1.7.1 h1:cuNEagBQEHWN1FnbGEjCXL2szYEXqfJPbP2HNUaca9Y=
+github.com/spf13/cast v1.10.0 h1:h2x0u2shc1QuLHfxi+cTJvs30+ZAHOGRic8uyGTDWxY=
-github.com/spf13/cast v1.7.1/go.mod h1:ancEpBxwJDODSW/UG4rDrAqiKolqNNh2DX3mk86cAdo=
+github.com/spf13/cast v1.10.0/go.mod h1:jNfB8QC9IA6ZuY2ZjDp0KtFO2LZZlg4S/7bzP6qqeHo=
-github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
+github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
-github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
+github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
-github.com/twmb/franz-go v1.18.1 h1:D75xxCDyvTqBSiImFx2lkPduE39jz1vaD7+FNc+vMkc=
+github.com/twmb/franz-go v1.19.5 h1:W7+o8D0RsQsedqib71OVlLeZ0zI6CbFra7yTYhZTs5Y=
-github.com/twmb/franz-go v1.18.1/go.mod h1:Uzo77TarcLTUZeLuGq+9lNpSkfZI+JErv7YJhlDjs9M=
+github.com/twmb/franz-go v1.19.5/go.mod h1:4kFJ5tmbbl7asgwAGVuyG1ZMx0NNpYk7EqflvWfPCpM=
-github.com/twmb/franz-go/pkg/kadm v1.16.0 h1:STMs1t5lYR5mR974PSiwNzE5TvsosByTp+rKXLOhAjE=
+github.com/twmb/franz-go/pkg/kadm v1.16.1 h1:IEkrhTljgLHJ0/hT/InhXGjPdmWfFvxp7o/MR7vJ8cw=
-github.com/twmb/franz-go/pkg/kadm v1.16.0/go.mod h1:MUdcUtnf9ph4SFBLLA/XxE29rvLhWYLM9Ygb8dfSCvw=
+github.com/twmb/franz-go/pkg/kadm v1.16.1/go.mod h1:Ue/ye1cc9ipsQFg7udFbbGiFNzQMqiH73fGC2y0rwyc=
+github.com/twmb/franz-go/pkg/kfake v0.0.0-20250508175730-72e1646135e3 h1:p24opKWPySAy8xSl8NqRgOv7Q+bX7kdrQirBVRJzQfo=
+github.com/twmb/franz-go/pkg/kfake v0.0.0-20250508175730-72e1646135e3/go.mod h1:7uQs3Ae6HkWT1Y9elMbqtAcNFCI0y6+iS+Phw49L49U=
 github.com/twmb/franz-go/pkg/kmsg v1.11.2 h1:hIw75FpwcAjgeyfIGFqivAvwC5uNIOWRGvQgZhH4mhg=
 github.com/twmb/franz-go/pkg/kmsg v1.11.2/go.mod h1:CFfkkLysDNmukPYhGzuUcDtf46gQSqCZHMW1T4Z+wDE=
-go.opentelemetry.io/otel v1.35.0 h1:xKWKPxrxB6OtMCbmMY021CqC45J+3Onta9MqjhnusiQ=
+go.opentelemetry.io/otel v1.38.0 h1:RkfdswUDRimDg0m2Az18RKOsnI8UDzppJAtj01/Ymk8=
-go.opentelemetry.io/otel v1.35.0/go.mod h1:UEqy8Zp11hpkUrL73gSlELM0DupHoiq72dR+Zqel/+Y=
+go.opentelemetry.io/otel v1.38.0/go.mod h1:zcmtmQ1+YmQM9wrNsTGV/q/uyusom3P8RxwExxkZhjM=
 go.unistack.org/micro-proto/v4 v4.1.0 h1:qPwL2n/oqh9RE3RTTDgt28XK3QzV597VugQPaw9lKUk=
 go.unistack.org/micro-proto/v4 v4.1.0/go.mod h1:ArmK7o+uFvxSY3dbJhKBBX4Pm1rhWdLEFf3LxBrMtec=
-go.unistack.org/micro/v4 v4.1.7 h1:4/dSEMTqxYoHimn/8wKDohfTXi2zDy6eWXYBCnBaZdc=
+go.unistack.org/micro/v4 v4.1.21 h1:F9PrbI1BhXSDS0FopwcO5wWrT2Xh38w9VhVZZBHjfg8=
-go.unistack.org/micro/v4 v4.1.7/go.mod h1:lr3oYED8Ay1vjK68QqRw30QOtdk/ffpZqMFDasOUhKw=
+go.unistack.org/micro/v4 v4.1.21/go.mod h1:nlBXTbx0rQrSZX4HPp2m57PHmpuGPWUd0O+jpUIiPto=
-golang.org/x/crypto v0.37.0 h1:kJNSjF/Xp7kU0iB2Z+9viTPMW4EqqsrywMXLJOOsXSE=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/crypto v0.37.0/go.mod h1:vg+k43peMZ0pUMhYmVAWysMK35e6ioLh3wB8ZCAfbVc=
+golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
-google.golang.org/protobuf v1.36.6 h1:z1NpPI8ku2WgiWnf+t9wTPsn6eP1L7ksHUlkfLvd9xY=
+google.golang.org/protobuf v1.36.10 h1:AYd7cD/uASjIL6Q9LiTjz8JLcrh/88q5UObnmY3aOOE=
-google.golang.org/protobuf v1.36.6/go.mod h1:jduwjTPXsFjZGTmRluh+L6NjiWu7pchiJ2/5YcXBHnY=
+google.golang.org/protobuf v1.36.10/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c/go.mod h1:JHkPIbrfpd72SG/EVd6muEfDQjcINNoR0C8j2r3qZ4Q=

hook_event.go

@@ -0,0 +1,108 @@
+package kgo
+
+import (
+	"context"
+	"net"
+	"sync/atomic"
+	"time"
+
+	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v4/logger"
+)
+
+type hookEvent struct {
+	log          logger.Logger
+	fatalOnError bool
+	connected    *atomic.Uint32
+}
+
+var (
+	_ kgo.HookBrokerConnect           = &hookEvent{}
+	_ kgo.HookBrokerDisconnect        = &hookEvent{}
+	_ kgo.HookBrokerRead              = &hookEvent{}
+	_ kgo.HookBrokerWrite             = &hookEvent{}
+	_ kgo.HookGroupManageError        = &hookEvent{}
+	_ kgo.HookProduceRecordUnbuffered = &hookEvent{}
+)
+
+func (m *hookEvent) OnGroupManageError(err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		ctx := context.TODO()
+		logMsg := "kgo.OnGroupManageError"
+
+		if m.fatalOnError {
+			m.log.Fatal(ctx, logMsg, err)
+		} else {
+			m.log.Error(ctx, logMsg, err)
+		}
+	}
+}
+
+func (m *hookEvent) OnBrokerConnect(_ kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		ctx := context.TODO()
+		logMsg := "kgo.OnBrokerConnect"
+
+		if m.fatalOnError {
+			m.log.Fatal(ctx, logMsg, err)
+		} else {
+			m.log.Error(ctx, logMsg, err)
+		}
+	}
+}
+
+func (m *hookEvent) OnBrokerDisconnect(_ kgo.BrokerMetadata, _ net.Conn) {}
+
+func (m *hookEvent) OnBrokerWrite(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		ctx := context.TODO()
+		logMsg := "kgo.OnBrokerWrite"
+
+		if m.fatalOnError {
+			m.log.Fatal(ctx, logMsg, err)
+		} else {
+			m.log.Error(ctx, logMsg, err)
+		}
+	}
+}
+
+func (m *hookEvent) OnBrokerRead(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		ctx := context.TODO()
+		logMsg := "kgo.OnBrokerRead"
+
+		if m.fatalOnError {
+			m.log.Fatal(ctx, logMsg, err)
+		} else {
+			m.log.Error(ctx, logMsg, err)
+		}
+	}
+}
+
+func (m *hookEvent) OnProduceRecordUnbuffered(_ *kgo.Record, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		ctx := context.TODO()
+		logMsg := "kgo.OnProduceRecordUnbuffered"
+
+		if m.fatalOnError {
+			m.log.Fatal(ctx, logMsg, err)
+		} else {
+			m.log.Error(ctx, logMsg, err)
+		}
+	}
+}

hook_event_test.go

@@ -0,0 +1,471 @@
+package kgo
+
+import (
+	"context"
+	"errors"
+	"io"
+	"net"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v4/logger"
+)
+
+func TestHookEvent_OnGroupManageError(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		fatalOnError          bool
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+		expectedFatalIsCalled bool
+		expectedFatalMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (non-fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          false,
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnGroupManageError",
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          true,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: true,
+			expectedFatalMsg:      "kgo.OnGroupManageError",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log, fatalOnError: tt.fatalOnError}
+			he.OnGroupManageError(tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+			require.Equal(t, tt.expectedFatalIsCalled, log.fatalIsCalled)
+			require.Equal(t, tt.expectedFatalMsg, log.fatalMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerConnect(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		fatalOnError          bool
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+		expectedFatalIsCalled bool
+		expectedFatalMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (non-fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          false,
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerConnect",
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          true,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: true,
+			expectedFatalMsg:      "kgo.OnBrokerConnect",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log, fatalOnError: tt.fatalOnError}
+			he.OnBrokerConnect(kgo.BrokerMetadata{}, 0, nil, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+			require.Equal(t, tt.expectedFatalIsCalled, log.fatalIsCalled)
+			require.Equal(t, tt.expectedFatalMsg, log.fatalMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerWrite(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		fatalOnError          bool
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+		expectedFatalIsCalled bool
+		expectedFatalMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (non-fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          false,
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerWrite",
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          true,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: true,
+			expectedFatalMsg:      "kgo.OnBrokerWrite",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log, fatalOnError: tt.fatalOnError}
+			he.OnBrokerWrite(kgo.BrokerMetadata{}, 0, 0, 0, 0, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+			require.Equal(t, tt.expectedFatalIsCalled, log.fatalIsCalled)
+			require.Equal(t, tt.expectedFatalMsg, log.fatalMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerRead(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		fatalOnError          bool
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+		expectedFatalIsCalled bool
+		expectedFatalMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (non-fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          false,
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerRead",
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          true,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: true,
+			expectedFatalMsg:      "kgo.OnBrokerRead",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log, fatalOnError: tt.fatalOnError}
+			he.OnBrokerRead(kgo.BrokerMetadata{}, 0, 0, 0, 0, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+			require.Equal(t, tt.expectedFatalIsCalled, log.fatalIsCalled)
+			require.Equal(t, tt.expectedFatalMsg, log.fatalMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnProduceRecordUnbuffered(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		fatalOnError          bool
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+		expectedFatalIsCalled bool
+		expectedFatalMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (non-fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          false,
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnProduceRecordUnbuffered",
+			expectedFatalIsCalled: false,
+		},
+		{
+			name:                  "some error (fatal)",
+			inputErr:              errors.New("some error"),
+			fatalOnError:          true,
+			expectedErrorIsCalled: false,
+			expectedFatalIsCalled: true,
+			expectedFatalMsg:      "kgo.OnProduceRecordUnbuffered",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log, fatalOnError: tt.fatalOnError}
+			he.OnProduceRecordUnbuffered(&kgo.Record{}, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+			require.Equal(t, tt.expectedFatalIsCalled, log.fatalIsCalled)
+			require.Equal(t, tt.expectedFatalMsg, log.fatalMsg)
+		})
+	}
+}
+
+// Mocks
+
+type mockLogger struct {
+	errorIsCalled bool
+	errorMsg      string
+
+	fatalIsCalled bool
+	fatalMsg      string
+}
+
+func (m *mockLogger) Init(...logger.Option) error {
+	panic("implement me")
+}
+
+func (m *mockLogger) Clone(...logger.Option) logger.Logger {
+	panic("implement me")
+}
+
+func (m *mockLogger) V(logger.Level) bool {
+	panic("implement me")
+}
+
+func (m *mockLogger) Level(logger.Level) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Options() logger.Options {
+	panic("implement me")
+}
+
+func (m *mockLogger) Fields(...interface{}) logger.Logger {
+	panic("implement me")
+}
+
+func (m *mockLogger) Info(context.Context, string, ...interface{}) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Trace(context.Context, string, ...interface{}) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Debug(context.Context, string, ...interface{}) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Warn(context.Context, string, ...interface{}) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Error(ctx context.Context, msg string, args ...interface{}) {
+	m.errorIsCalled = true
+	m.errorMsg = msg
+}
+
+func (m *mockLogger) Fatal(ctx context.Context, msg string, args ...interface{}) {
+	m.fatalIsCalled = true
+	m.fatalMsg = msg
+}
+
+func (m *mockLogger) Log(context.Context, logger.Level, string, ...interface{}) {
+	panic("implement me")
+}
+
+func (m *mockLogger) Name() string {
+	panic("implement me")
+}
+
+func (m *mockLogger) String() string {
+	panic("implement me")
+}

kgo.go

@@ -6,11 +6,13 @@ import (
 	"errors"
 	"fmt"
 	"math/rand/v2"
+	"strconv"
 	"strings"
 	"sync"
 	"sync/atomic"
 	"time"
 
+	"github.com/twmb/franz-go/pkg/kadm"
 	"github.com/twmb/franz-go/pkg/kgo"
 	"github.com/twmb/franz-go/pkg/kmsg"
 	"go.unistack.org/micro/v4/broker"
@@ -74,7 +76,7 @@ type Broker struct {
 
 	opts broker.Options
 
-	sync.RWMutex
+	mu   sync.RWMutex
 	init bool
 }
 
@@ -108,7 +110,7 @@ type kgoMessage struct {
 	ctx   context.Context
 	body  []byte
 	hdr   metadata.Metadata
-	opts  broker.PublishOptions
+	opts  broker.MessageOptions
 	ack   bool
 }
 
@@ -141,18 +143,22 @@ func (b *Broker) newCodec(ct string) (codec.Codec, error) {
 	if idx := strings.IndexRune(ct, ';'); idx >= 0 {
 		ct = ct[:idx]
 	}
-	b.RLock()
+	b.mu.RLock()
 	c, ok := b.opts.Codecs[ct]
-	b.RUnlock()
+	b.mu.RUnlock()
 	if ok {
 		return c, nil
 	}
 	return nil, codec.ErrUnknownContentType
 }
 
-func (b *Broker) NewMessage(ctx context.Context, hdr metadata.Metadata, body interface{}, opts ...broker.PublishOption) (broker.Message, error) {
+func (b *Broker) NewMessage(ctx context.Context, hdr metadata.Metadata, body interface{}, opts ...broker.MessageOption) (broker.Message, error) {
-	options := broker.NewPublishOptions(opts...)
+	options := broker.NewMessageOptions(opts...)
-	m := &kgoMessage{ctx: ctx, hdr: hdr, opts: options}
+	if options.ContentType == "" {
+		options.ContentType = b.opts.ContentType
+	}
+
+	m := &kgoMessage{ctx: ctx, hdr: hdr.Copy(), opts: options}
 	c, err := b.newCodec(m.opts.ContentType)
 	if err == nil {
 		m.body, err = c.Marshal(body)
@@ -161,6 +167,8 @@ func (b *Broker) NewMessage(ctx context.Context, hdr metadata.Metadata, body int
 		return nil, err
 	}
 
+	m.hdr.Set(metadata.HeaderContentType, m.opts.ContentType)
+
 	return m, nil
 }
 
@@ -234,10 +242,72 @@ func (k *Broker) Connect(ctx context.Context) error {
 		return err
 	}
 
-	k.Lock()
+	k.mu.Lock()
 	k.c = c
 	k.connected.Store(1)
-	k.Unlock()
+	k.mu.Unlock()
+
+	exposeLag := false
+	if k.opts.Context != nil {
+		if v, ok := k.opts.Context.Value(exposeLagKey{}).(bool); ok && v {
+			exposeLag = v
+		}
+	}
+
+	if exposeLag {
+		var mu sync.Mutex
+		var lastUpdate time.Time
+		type pl struct {
+			p string
+			l float64
+		}
+
+		lag := make(map[string]map[string]pl) // topic => group => partition => lag
+		ac := kadm.NewClient(k.c)
+
+		updateStats := func() {
+			mu.Lock()
+			if time.Since(lastUpdate) < DefaultStatsInterval {
+				return
+			}
+			mu.Unlock()
+
+			k.mu.Lock()
+			groups := make([]string, 0, len(k.subs))
+			for _, g := range k.subs {
+				groups = append(groups, g.opts.Group)
+			}
+			k.mu.Unlock()
+
+			dgls, err := ac.Lag(ctx, groups...)
+			if err != nil || !dgls.Ok() {
+				k.opts.Logger.Error(k.opts.Context, "kgo describe group lag error", err)
+				return
+			}
+
+			for gn, dgl := range dgls {
+				for tn, lmap := range dgl.Lag {
+					if _, ok := lag[tn]; !ok {
+						lag[tn] = make(map[string]pl)
+					}
+					for p, l := range lmap {
+						lag[tn][gn] = pl{p: strconv.Itoa(int(p)), l: float64(l.Lag)}
+					}
+				}
+			}
+		}
+
+		for tn, dg := range lag {
+			for gn, gl := range dg {
+				k.opts.Meter.Gauge(semconv.BrokerGroupLag,
+					func() float64 { updateStats(); return gl.l },
+					"topic", tn,
+					"group", gn,
+					"partition", gl.p)
+			}
+		}
+
+	}
 
 	return nil
 }
@@ -255,8 +325,8 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 	ctx, span = k.opts.Tracer.Start(ctx, "Disconnect")
 	defer span.Finish()
 
-	k.Lock()
+	k.mu.Lock()
-	defer k.Unlock()
+	defer k.mu.Unlock()
 	select {
 	case <-nctx.Done():
 		return nctx.Err()
@@ -280,8 +350,8 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 }
 
 func (k *Broker) Init(opts ...broker.Option) error {
-	k.Lock()
+	k.mu.Lock()
-	defer k.Unlock()
+	defer k.mu.Unlock()
 
 	if len(opts) == 0 && k.init {
 		return nil
@@ -340,51 +410,44 @@ func (b *Broker) fnPublish(ctx context.Context, topic string, messages ...broker
 }
 
 func (b *Broker) publish(ctx context.Context, topic string, messages ...broker.Message) error {
-	if b.connected.Load() == 0 {
+	var records []*kgo.Record
-		c, _, err := b.connect(ctx, b.kopts...)
-		if err != nil {
-			return err
-		}
-		b.Lock()
-		b.c = c
-		b.Unlock()
-	}
-
-	records := make([]*kgo.Record, 0, len(messages))
-	var errs []string
-	var key []byte
-	var promise func(*kgo.Record, error)
 
 	for _, msg := range messages {
 
-		if mctx := msg.Context(); mctx != nil {
-			if k, ok := mctx.Value(publishKey{}).([]byte); ok && k != nil {
-				key = k
-			}
-			if p, ok := mctx.Value(publishPromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
-				promise = p
-			}
-		}
-
 		rec := &kgo.Record{
-			Context: ctx,
+			Context: msg.Context(),
-			Key:     key,
 			Topic:   topic,
 			Value:   msg.Body(),
 		}
 
-		b.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Inc()
+		var promise func(*kgo.Record, error)
+		if rec.Context != nil {
+			if k, ok := rec.Context.Value(messageKey{}).([]byte); ok && k != nil {
+				rec.Key = k
+			}
+			if p, ok := rec.Context.Value(messagePromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
+				promise = p
+			}
+		}
+
+		kmsg, ok := msg.(*kgoMessage)
+		if !ok {
+			continue
+		}
+		if kmsg.opts.Context != nil {
+			if k, ok := kmsg.opts.Context.Value(messageKey{}).([]byte); ok && k != nil {
+				rec.Key = k
+			}
+			if p, ok := kmsg.opts.Context.Value(messagePromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
+				promise = p
+			}
+		}
 
 		setHeaders(rec, msg.Header())
 
-		records = append(records, rec)
-	}
-
-	ts := time.Now()
-
 		if promise != nil {
-
+			ts := time.Now()
-		for _, rec := range records {
+			b.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Inc()
 			b.c.Produce(ctx, rec, func(r *kgo.Record, err error) {
 				te := time.Since(ts)
 				b.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Dec()
@@ -397,12 +460,17 @@ func (b *Broker) publish(ctx context.Context, topic string, messages ...broker.M
 				}
 				promise(r, err)
 			})
+			continue
+		} else {
+			records = append(records, rec)
 		}
-		return nil
 	}
 
+	if len(records) > 0 {
+		var errs []string
+		ts := time.Now()
+		b.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", topic, "topic", topic).Set(uint64(len(records)))
 		results := b.c.ProduceSync(ctx, records...)
-
 		te := time.Since(ts)
 		for _, result := range results {
 			b.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
@@ -419,6 +487,7 @@ func (b *Broker) publish(ctx context.Context, topic string, messages ...broker.M
 		if len(errs) > 0 {
 			return fmt.Errorf("publish error: %s", strings.Join(errs, "\n"))
 		}
+	}
 
 	return nil
 }
@@ -474,11 +543,15 @@ func (b *Broker) fnSubscribe(ctx context.Context, topic string, handler interfac
 		}
 	}
 
+	var messagePool bool
 	var fatalOnError bool
 	if b.opts.Context != nil {
 		if v, ok := b.opts.Context.Value(fatalOnErrorKey{}).(bool); ok && v {
 			fatalOnError = v
 		}
+		if v, ok := b.opts.Context.Value(subscribeMessagePoolKey{}).(bool); ok && v {
+			messagePool = v
+		}
 	}
 
 	if options.Context != nil {
@@ -496,6 +569,7 @@ func (b *Broker) fnSubscribe(ctx context.Context, topic string, handler interfac
 		done:         make(chan struct{}),
 		fatalOnError: fatalOnError,
 		connected:    b.connected,
+		messagePool:  messagePool,
 	}
 
 	kopts := append(b.kopts,
@@ -540,9 +614,9 @@ func (b *Broker) fnSubscribe(ctx context.Context, topic string, handler interfac
 
 	go sub.poll(ctx)
 
-	b.Lock()
+	b.mu.Lock()
 	b.subs = append(b.subs, sub)
-	b.Unlock()
+	b.mu.Unlock()
 
 	return sub, nil
 }

kgo_test.go

@@ -2,15 +2,15 @@ package kgo_test
 
 import (
 	"context"
-	"os"
-	"strings"
 	"sync/atomic"
 	"testing"
 	"time"
 
+	"github.com/twmb/franz-go/pkg/kfake"
 	kg "github.com/twmb/franz-go/pkg/kgo"
 	kgo "go.unistack.org/micro-broker-kgo/v4"
 	"go.unistack.org/micro/v4/broker"
+	"go.unistack.org/micro/v4/codec"
 	"go.unistack.org/micro/v4/logger"
 	"go.unistack.org/micro/v4/logger/slog"
 	"go.unistack.org/micro/v4/metadata"
@@ -20,31 +20,33 @@ var (
 	msgcnt   = int64(1200)
 	group    = "38"
 	prefill  = true
-	loglevel = logger.DebugLevel
+	loglevel = logger.ErrorLevel
+	cluster  *kfake.Cluster
 )
 
-func TestFail(t *testing.T) {
+func TestMain(m *testing.M) {
-	if tr := os.Getenv("INTEGRATION_TESTS"); len(tr) > 0 {
+	cluster = kfake.MustCluster(
-		t.Skip()
+		kfake.AllowAutoTopicCreation(),
-	}
+	)
+	defer cluster.Close()
+	m.Run()
+}
 
+func TestFail(t *testing.T) {
 	logger.DefaultLogger = slog.NewLogger()
 	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel)); err != nil {
 		t.Fatal(err)
 	}
 	ctx := context.Background()
 
-	var addrs []string
-	if addr := os.Getenv("BROKER_ADDRS"); len(addr) == 0 {
-		addrs = []string{"127.0.0.1:9092"}
-	} else {
-		addrs = strings.Split(addr, ",")
-	}
-
 	b := kgo.NewBroker(
-		broker.Addrs(addrs...),
+		broker.ContentType("application/octet-stream"),
+		broker.Codec("application/octet-stream", codec.NewCodec()),
+		broker.Addrs(cluster.ListenAddrs()...),
 		kgo.CommitInterval(5*time.Second),
-		kgo.Options(kg.ClientID("test"), kg.FetchMaxBytes(10*1024*1024),
+		kgo.Options(
+			kg.ClientID("test"),
+			kg.FetchMaxBytes(10*1024*1024),
 			kg.AllowAutoTopicCreation(),
 		),
 	)
@@ -69,33 +71,39 @@ func TestFail(t *testing.T) {
 	t.Logf("broker health %v", b.Health())
 	msgs := make([]broker.Message, 0, msgcnt)
 	for i := int64(0); i < msgcnt; i++ {
-		m, _ := b.NewMessage(ctx, metadata.Pairs("hkey", "hval"), []byte(`test`))
+		m, err := b.NewMessage(ctx, metadata.Pairs("hkey", "hval"), []byte(`test`))
+		if err != nil {
+			t.Fatal(err)
+		}
 		msgs = append(msgs, m)
 	}
 
+	go func() {
 		for _, msg := range msgs {
-		t.Logf("broker publish")
+			//		t.Logf("broker publish")
-		if err := b.Publish(ctx, "test", msg); err != nil {
+			if err := b.Publish(ctx, "test.fail", msg); err != nil {
 				t.Fatal(err)
 			}
 		}
+	}()
 	//	t.Skip()
 
 	idx := int64(0)
 	fn := func(msg broker.Message) error {
 		atomic.AddInt64(&idx, 1)
-		time.Sleep(500 * time.Millisecond)
+		time.Sleep(100 * time.Millisecond)
-		t.Logf("ack")
+		// t.Logf("ack")
 		return msg.Ack()
 	}
 
-	sub, err := b.Subscribe(ctx, "test", fn,
+	sub, err := b.Subscribe(ctx, "test.fail", fn,
 		broker.SubscribeAutoAck(true),
 		broker.SubscribeGroup(group),
 		broker.SubscribeBodyOnly(true))
 	if err != nil {
 		t.Fatal(err)
 	}
+
 	defer func() {
 		if err := sub.Unsubscribe(ctx); err != nil {
 			t.Fatal(err)
@@ -108,17 +116,26 @@ func TestFail(t *testing.T) {
 			t.Logf("health works")
 			break
 		}
-		time.Sleep(1 * time.Second)
+		t.Logf("health sleep")
+		time.Sleep(100 * time.Millisecond)
+		if err := b.Disconnect(ctx); err != nil {
+			t.Fatal(err)
+		}
 	}
 }
 
 func TestConnect(t *testing.T) {
-	var addrs []string
 	ctx := context.TODO()
 	b := kgo.NewBroker(
-		broker.Addrs(addrs...),
+		broker.ContentType("application/octet-stream"),
+		broker.Codec("application/octet-stream", codec.NewCodec()),
+		broker.Addrs(cluster.ListenAddrs()...),
 		kgo.CommitInterval(5*time.Second),
-		kgo.Options(kg.ClientID("test"), kg.FetchMaxBytes(10*1024*1024)),
+		kgo.Options(
+			kg.ClientID("test"),
+			kg.FetchMaxBytes(10*1024*1024),
+			kg.AllowAutoTopicCreation(),
+		),
 	)
 	if err := b.Init(); err != nil {
 		t.Fatal(err)
@@ -130,26 +147,21 @@ func TestConnect(t *testing.T) {
 }
 
 func TestPubSub(t *testing.T) {
-	if tr := os.Getenv("INTEGRATION_TESTS"); len(tr) > 0 {
-		t.Skip()
-	}
-
 	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel)); err != nil {
 		t.Fatal(err)
 	}
 	ctx := context.Background()
 
-	var addrs []string
-	if addr := os.Getenv("BROKER_ADDRS"); len(addr) == 0 {
-		addrs = []string{"127.0.0.1:29091", "127.0.0.2:29092", "127.0.0.3:29093"}
-	} else {
-		addrs = strings.Split(addr, ",")
-	}
-
 	b := kgo.NewBroker(
-		broker.Addrs(addrs...),
+		broker.ContentType("application/octet-stream"),
+		broker.Codec("application/octet-stream", codec.NewCodec()),
+		broker.Addrs(cluster.ListenAddrs()...),
 		kgo.CommitInterval(5*time.Second),
-		kgo.Options(kg.ClientID("test"), kg.FetchMaxBytes(10*1024*1024)),
+		kgo.Options(
+			kg.ClientID("test"),
+			kg.FetchMaxBytes(10*1024*1024),
+			kg.AllowAutoTopicCreation(),
+		),
 	)
 
 	if err := b.Init(); err != nil {
@@ -172,7 +184,7 @@ func TestPubSub(t *testing.T) {
 			msgs = append(msgs, m)
 		}
 
-		if err := b.Publish(ctx, "test", msgs...); err != nil {
+		if err := b.Publish(ctx, "test.pubsub", msgs...); err != nil {
 			t.Fatal(err)
 		}
 		//	t.Skip()
@@ -185,7 +197,7 @@ func TestPubSub(t *testing.T) {
 		return msg.Ack()
 	}
 
-	sub, err := b.Subscribe(ctx, "test", fn,
+	sub, err := b.Subscribe(ctx, "test.pubsub", fn,
 		broker.SubscribeAutoAck(true),
 		broker.SubscribeGroup(group),
 		broker.SubscribeBodyOnly(true))
@@ -210,7 +222,7 @@ func TestPubSub(t *testing.T) {
 				if prc := atomic.LoadInt64(&idx); prc == msgcnt {
 					close(done)
 				} else {
-					t.Logf("processed %v\n", prc)
+					t.Logf("processed %v of %v\n", prc, msgcnt)
 				}
 			case <-ticker.C:
 				close(done)

options.go

@@ -14,7 +14,7 @@ var (
 	DefaultCommitInterval = 5 * time.Second
 
 	// DefaultStatsInterval specifies how fast check consumer lag
-	DefaultStatsInterval = 5 * time.Second
+	DefaultStatsInterval = 30 * time.Second
 
 	// DefaultSubscribeMaxInflight specifies how much messages keep inflight
 	DefaultSubscribeMaxInflight = 100
@@ -27,11 +27,11 @@ func SubscribeContext(ctx context.Context) broker.SubscribeOption {
 	return broker.SetSubscribeOption(subscribeContextKey{}, ctx)
 }
 
-type publishKey struct{}
+type messageKey struct{}
 
-// PublishKey set the kafka message key (broker option)
+// MessageKey set the kafka message key (broker option)
-func PublishKey(key []byte) broker.PublishOption {
+func MessageKey(key []byte) broker.MessageOption {
-	return broker.SetPublishOption(publishKey{}, key)
+	return broker.SetMessageOption(messageKey{}, key)
 }
 
 type optionsKey struct{}
@@ -103,9 +103,23 @@ func SubscribeFatalOnError(b bool) broker.SubscribeOption {
 	return broker.SetSubscribeOption(fatalOnErrorKey{}, b)
 }
 
-type publishPromiseKey struct{}
+type messagePromiseKey struct{}
 
-// PublishPromise set the kafka promise func for Produce
+// MessagePromise set the kafka promise func for Produce
-func PublishPromise(fn func(*kgo.Record, error)) broker.PublishOption {
+func MessagePromise(fn func(*kgo.Record, error)) broker.MessageOption {
-	return broker.SetPublishOption(publishPromiseKey{}, fn)
+	return broker.SetMessageOption(messagePromiseKey{}, fn)
+}
+
+type subscribeMessagePoolKey struct{}
+
+// SubscribeMessagePool optionaly enabled/disable message pool
+func SubscribeMessagePool(b bool) broker.SubscribeOption {
+	return broker.SetSubscribeOption(subscribeMessagePoolKey{}, b)
+}
+
+type exposeLagKey struct{}
+
+// ExposeLag enabled subscriber lag via [meter.Meter]
+func ExposeLag(b bool) broker.Option {
+	return broker.SetOption(exposeLagKey{}, b)
 }

subscriber.go

@@ -8,7 +8,6 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/twmb/franz-go/pkg/kadm"
 	"github.com/twmb/franz-go/pkg/kgo"
 	"github.com/twmb/franz-go/pkg/kmsg"
 	"go.unistack.org/micro/v4/broker"
@@ -35,6 +34,7 @@ type consumer struct {
 	opts        broker.SubscribeOptions
 	handler     interface{}
 	connected   *atomic.Uint32
+	messagePool bool
 }
 
 type Subscriber struct {
@@ -42,14 +42,13 @@ type Subscriber struct {
 	c            *kgo.Client
 	htracer      *hookTracer
 	topic        string
-
+	messagePool  bool
 	handler      interface{}
 	done         chan struct{}
 	kopts        broker.Options
 	opts         broker.SubscribeOptions
-
 	connected    *atomic.Uint32
-	sync.RWMutex
+	mu           sync.RWMutex
 	closed       bool
 	fatalOnError bool
 }
@@ -87,46 +86,13 @@ func (s *Subscriber) Unsubscribe(ctx context.Context) error {
 
 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))).Set(uint64(l.Lag))
-				}
-				s.Unlock()
-
-			}
-		}
-	}()
-
 	for {
 		select {
 		case <-ctx.Done():
@@ -141,12 +107,21 @@ func (s *Subscriber) poll(ctx context.Context) {
 				return
 			}
 			fetches.EachError(func(t string, p int32, err error) {
+				if kgo.IsRetryableBrokerErr(err) {
+					s.kopts.Logger.Error(ctx, fmt.Sprintf("[kgo] fetch topic %s partition %d error", t, p), err)
+				} else {
 					s.kopts.Logger.Fatal(ctx, fmt.Sprintf("[kgo] fetch topic %s partition %d error", t, p), err)
+				}
 			})
 
 			fetches.EachPartition(func(p kgo.FetchTopicPartition) {
 				tps := tp{p.Topic, p.Partition}
-				s.consumers[tps].recs <- p
+				s.mu.Lock()
+				c := s.consumers[tps]
+				s.mu.Unlock()
+				if c != nil {
+					c.recs <- p
+				}
 			})
 			s.c.AllowRebalance()
 		}
@@ -160,11 +135,15 @@ func (s *Subscriber) killConsumers(ctx context.Context, lost map[string][]int32)
 	for topic, partitions := range lost {
 		for _, partition := range partitions {
 			tps := tp{topic, partition}
+			s.mu.Lock()
 			pc, ok := s.consumers[tps]
+			s.mu.Unlock()
 			if !ok {
 				continue
 			}
+			s.mu.Lock()
 			delete(s.consumers, tps)
+			s.mu.Unlock()
 			close(pc.quit)
 			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))
@@ -215,13 +194,14 @@ func (s *Subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[str
 				done:        make(chan struct{}),
 				recs:        make(chan kgo.FetchTopicPartition, 100),
 				handler:     s.handler,
+				messagePool: s.messagePool,
 				kopts:       s.kopts,
 				opts:        s.opts,
 				connected:   s.connected,
 			}
-			s.Lock()
+			s.mu.Lock()
 			s.consumers[tp{topic, partition}] = pc
-			s.Unlock()
+			s.mu.Unlock()
 			go pc.consume()
 		}
 	}
@@ -236,6 +216,8 @@ func (pc *consumer) consume() {
 		defer pc.kopts.Logger.Debug(pc.kopts.Context, fmt.Sprintf("killing, topic %s partition %d", pc.topic, pc.partition))
 	}
 
+	var pm *kgoMessage
+
 	for {
 		select {
 		case <-pc.quit:
@@ -245,50 +227,64 @@ func (pc *consumer) consume() {
 				ctx, sp := pc.htracer.WithProcessSpan(record)
 				ts := time.Now()
 				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Inc()
-				p := messagePool.Get().(*kgoMessage)
 
-				p.body = record.Value
+				if pc.messagePool {
-				p.topic = record.Topic
+					pm = messagePool.Get().(*kgoMessage)
-				p.ack = false
+				} else {
-				p.hdr = metadata.New(len(record.Headers))
+					pm = &kgoMessage{}
-				p.ctx = ctx
-				for _, hdr := range record.Headers {
-					p.hdr.Set(hdr.Key, string(hdr.Value))
 				}
-
+				pm.body = record.Value
+				pm.topic = record.Topic
+				pm.ack = false
+				pm.hdr = metadata.New(len(record.Headers))
+				pm.ctx = ctx
+				for _, hdr := range record.Headers {
+					pm.hdr.Set(hdr.Key, string(hdr.Value))
+				}
+				pm.hdr.Set("Micro-Offset", strconv.FormatInt(record.Offset, 10))
+				pm.hdr.Set("Micro-Partition", strconv.FormatInt(int64(record.Partition), 10))
+				pm.hdr.Set("Micro-Topic", record.Topic)
+				pm.hdr.Set("Micro-Key", string(record.Key))
+				pm.hdr.Set("Micro-Timestamp", strconv.FormatInt(record.Timestamp.Unix(), 10))
 				switch h := pc.handler.(type) {
 				case func(broker.Message) error:
-					err = h(p)
+					err = h(pm)
 				case func([]broker.Message) error:
-					err = h([]broker.Message{p})
+					err = h([]broker.Message{pm})
 				}
 
 				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 				if err != nil {
+					if sp != nil {
 						sp.SetStatus(tracer.SpanStatusError, err.Error())
+					}
 					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "failure").Inc()
 				} else if pc.opts.AutoAck {
-					p.ack = true
+					pm.ack = true
 				}
 
 				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())
 
-				ack := p.ack
+				ack := pm.ack
+				if pc.messagePool {
 					messagePool.Put(p)
-
+				}
 				if ack {
 					pc.c.MarkCommitRecords(record)
 				} else {
+					if sp != nil {
 						sp.Finish()
+					}
 					//					pc.connected.Store(0)
 					pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] message not commited")
 					return
 				}
-
+				if sp != nil {
 					sp.Finish()
 				}
 			}
 		}
+	}
 }

tracer.go

@@ -32,6 +32,9 @@ var (
 // the record's context, so it can be ended in the OnProduceRecordUnbuffered
 // hook.
 func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	// Set up span options.
 	attrs := []interface{}{
 		messagingSystem,
@@ -77,6 +80,9 @@ 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) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	if span, ok := tracer.SpanFromContext(r.Context); ok {
 		span.AddLabels(
 			semconv.MessagingKafkaDestinationPartition(int(r.Partition)),
@@ -96,6 +102,9 @@ func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
 // OnFetchRecordUnbuffered hook and can be used in downstream consumer
 // processing.
 func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	// Set up the span options.
 	attrs := []interface{}{
 		messagingSystem,
@@ -141,6 +150,9 @@ func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
 // OnFetchRecordUnbuffered continues and ends the "receive" span for an
 // unbuffered record.
 func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	span, _ := tracer.SpanFromContext(r.Context)
 	span.Finish()
 }
@@ -155,6 +167,13 @@ func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
 // 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) {
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	if !m.tracer.Enabled() {
+		return r.Context, nil
+	}
 	// Set up the span options.
 	attrs := []interface{}{
 		messagingSystem,