update deps

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

.github/dependabot.yml

@@ -1,19 +0,0 @@
-# To get started with Dependabot version updates, you'll need to specify which
-# package ecosystems to update and where the package manifests are located.
-# Please see the documentation for all configuration options:
-# https://help.github.com/github/administering-a-repository/configuration-options-for-dependency-updates
-
-version: 2
-updates:
-
-  # Maintain dependencies for GitHub Actions
-  - package-ecosystem: "github-actions"
-    directory: "/"
-    schedule:
-      interval: "daily"
-
-  # Maintain dependencies for Golang
-  - package-ecosystem: "gomod"
-    directory: "/"
-    schedule:
-      interval: "daily"

.github/workflows/autoapprove.yml

@@ -1,20 +0,0 @@
-name: "autoapprove"
-
-on:
-  pull_request_target:
-    types: [assigned, opened, synchronize, reopened]
-
-permissions:
-  pull-requests: write
-  contents: write
-
-jobs:
-  autoapprove:
-    runs-on: ubuntu-latest
-    steps:
-      - name: approve
-        uses: hmarr/auto-approve-action@v3
-        if: github.actor == 'vtolstov' || github.actor == 'dependabot[bot]'
-        id: approve
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/automerge.yml

@@ -1,21 +0,0 @@
-name: "automerge"
-
-on:
-  pull_request_target:
-    types: [assigned, opened, synchronize, reopened]
-
-permissions:
-  pull-requests: write
-  contents: write
-
-jobs:
-  automerge:
-    runs-on: ubuntu-latest
-    if: github.actor == 'vtolstov'
-    steps:
-      - name: merge
-        id: merge
-        run: gh pr merge --auto --merge "$PR_URL"
-        env:
-          PR_URL: ${{github.event.pull_request.html_url}}
-          GITHUB_TOKEN: ${{secrets.TOKEN}}

.github/workflows/build.yml

@@ -1,47 +0,0 @@
-name: build
-on:
- push:
-    branches:
-    - master
-    - v3
-jobs:
-  test:
-    name: test
-    runs-on: ubuntu-latest
-    steps:
-    - name: setup
-      uses: actions/setup-go@v3
-      with:
-        go-version: 1.17
-    - name: checkout
-      uses: actions/checkout@v3
-    - name: cache
-      uses: actions/cache@v3
-      with:
-        path: ~/go/pkg/mod
-        key: ${{ runner.os }}-go-${{ hashFiles('**/go.sum') }}
-        restore-keys: ${{ runner.os }}-go-
-    - name: deps
-      run: go get -v -t -d ./...
-    - name: test
-      env:
-        INTEGRATION_TESTS: yes
-      run: go test -mod readonly -v ./...
-  lint:
-    name: lint
-    runs-on: ubuntu-latest
-    steps:
-      - name: checkout
-        uses: actions/checkout@v3
-      - name: lint
-        uses: golangci/golangci-lint-action@v3.4.0
-        continue-on-error: true
-        with:
-          # Required: the version of golangci-lint is required and must be specified without patch version: we always use the latest patch version.
-          version: v1.30
-          # Optional: working directory, useful for monorepos
-          # working-directory: somedir
-          # Optional: golangci-lint command line arguments.
-          # args: --issues-exit-code=0
-          # Optional: show only new issues if it's a pull request. The default value is `false`.
-          # only-new-issues: true

.github/workflows/codeql-analysis.yml

@@ -1,78 +0,0 @@
-# For most projects, this workflow file will not need changing; you simply need
-# to commit it to your repository.
-#
-# You may wish to alter this file to override the set of languages analyzed,
-# or to provide custom queries or build logic.
-#
-# ******** NOTE ********
-# We have attempted to detect the languages in your repository. Please check
-# the `language` matrix defined below to confirm you have the correct set of
-# supported CodeQL languages.
-#
-name: "codeql"
-
-on:
-  workflow_run:
-    workflows: ["prbuild"]
-    types:
-      - completed
-  push:
-    branches: [ master, v3 ]
-  pull_request:
-    # The branches below must be a subset of the branches above
-    branches: [ master, v3 ]
-  schedule:
-    - cron: '34 1 * * 0'
-
-jobs:
-  analyze:
-    name: analyze
-    runs-on: ubuntu-latest
-    permissions:
-      actions: read
-      contents: read
-      security-events: write
-
-    strategy:
-      fail-fast: false
-      matrix:
-        language: [ 'go' ]
-        # CodeQL supports [ 'cpp', 'csharp', 'go', 'java', 'javascript', 'python' ]
-        # Learn more:
-        # https://docs.github.com/en/free-pro-team@latest/github/finding-security-vulnerabilities-and-errors-in-your-code/configuring-code-scanning#changing-the-languages-that-are-analyzed
-
-    steps:
-    - name: checkout
-      uses: actions/checkout@v3
-    - name: setup
-      uses: actions/setup-go@v3
-      with:
-        go-version: 1.17
-    # Initializes the CodeQL tools for scanning.
-    - name: init
-      uses: github/codeql-action/init@v2
-      with:
-        languages: ${{ matrix.language }}
-        # If you wish to specify custom queries, you can do so here or in a config file.
-        # By default, queries listed here will override any specified in a config file.
-        # Prefix the list here with "+" to use these queries and those in the config file.
-        # queries: ./path/to/local/query, your-org/your-repo/queries@main
-
-    # Autobuild attempts to build any compiled languages  (C/C++, C#, or Java).
-    # If this step fails, then you should remove it and run the build manually (see below)
-    - name: autobuild
-      uses: github/codeql-action/autobuild@v2
-
-    # ℹ️ Command-line programs to run using the OS shell.
-    # 📚 https://git.io/JvXDl
-
-    # ✏️ If the Autobuild fails above, remove it and uncomment the following three lines
-    #    and modify them (or add more) to build your code if your project
-    #    uses a compiled language
-
-    #- run: |
-    #   make bootstrap
-    #   make release
-
-    - name: analyze
-      uses: github/codeql-action/analyze@v2

.github/workflows/dependabot-automerge.yml

@@ -1,27 +0,0 @@
-name: "dependabot-automerge"
-
-on:
-  pull_request_target:
-    types: [assigned, opened, synchronize, reopened]
-
-permissions:
-  pull-requests: write
-  contents: write
-
-jobs:
-  automerge:
-    runs-on: ubuntu-latest
-    if: github.actor == 'dependabot[bot]'
-    steps:
-      - name: metadata
-        id: metadata
-        uses: dependabot/fetch-metadata@v1.3.6
-        with:
-          github-token: "${{ secrets.TOKEN }}"
-      - name: merge
-        id: merge
-        if: ${{contains(steps.metadata.outputs.dependency-names, 'go.unistack.org')}}
-        run: gh pr merge --auto --merge "$PR_URL"
-        env:
-          PR_URL: ${{github.event.pull_request.html_url}}
-          GITHUB_TOKEN: ${{secrets.TOKEN}}

.github/workflows/job_coverage.yml

@@ -0,0 +1,53 @@
+name: coverage
+
+on:
+  push:
+    branches: [ main, v3, v4 ]
+    paths-ignore:
+      - '.github/**'
+      - '.gitea/**'
+  pull_request:
+    branches: [ main, v3, v4 ]
+
+jobs:
+
+  build:
+    if: github.server_url != 'https://github.com'
+    runs-on: ubuntu-latest
+    steps:
+    - name: checkout code
+      uses: actions/checkout@v4
+      with:
+        filter: 'blob:none'
+
+    - name: setup go
+      uses: actions/setup-go@v5
+      with:
+        cache-dependency-path: "**/*.sum"
+        go-version: 'stable'
+
+    - name: test coverage
+      run: |
+        go test -v -cover ./... -covermode=count -coverprofile coverage.out -coverpkg ./...
+        go tool cover -func coverage.out -o coverage.out
+
+    - name: coverage badge
+      uses: tj-actions/coverage-badge-go@v2
+      with:
+        green: 80
+        filename: coverage.out
+
+    - uses: stefanzweifel/git-auto-commit-action@v4
+      name: autocommit
+      with:
+        commit_message: Apply Code Coverage Badge
+        skip_fetch: false
+        skip_checkout: false
+        file_pattern: ./README.md
+
+    - name: push
+      if: steps.auto-commit-action.outputs.changes_detected == 'true'
+      uses: ad-m/github-push-action@master
+      with:
+        github_token: ${{ github.token }}
+        branch: ${{ github.ref }}

.github/workflows/job_lint.yml

@@ -0,0 +1,29 @@
+name: lint
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches: [ master, v3, v4 ]
+    paths-ignore:
+      - '.github/**'
+      - '.gitea/**'
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+    - name: checkout code
+      uses: actions/checkout@v4
+      with:
+        filter: 'blob:none'
+    - name: setup go
+      uses: actions/setup-go@v5
+      with:
+        cache-dependency-path: "**/*.sum"
+        go-version: 'stable'
+    - name: setup deps
+      run: go get -v ./...
+    - name: run lint
+      uses: golangci/golangci-lint-action@v6
+      with:
+        version: 'latest'

.github/workflows/job_sync.yml

@@ -0,0 +1,94 @@
+name: sync
+
+on:
+  schedule:
+    - cron: '*/5 * * * *'
+  # Allows you to run this workflow manually from the Actions tab
+  workflow_dispatch:
+
+jobs:
+  sync:
+    if: github.server_url != 'https://github.com'
+    runs-on: ubuntu-latest
+    steps:
+    - name: init
+      run: |
+        git config --global user.email "vtolstov <vtolstov@users.noreply.github.com>"
+        git config --global user.name "github-actions[bot]"
+        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" -a "$src_hash" != "" -a "" != "$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/${GITHUB_REPOSITORY}
+        git pull --rebase upstream master
+        git push upstream master --progress
+        git push origin master --progress
+        cd ../
+        rm -rf repo
+
+    - name: check v3
+      id: check_v3
+      run: |
+        src_hash=$(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/${GITHUB_REPOSITORY}
+        git pull --rebase upstream v3
+        git push upstream v3 --progress
+        git push origin v3 --progress
+        cd ../
+        rm -rf repo
+
+    - name: check v4
+      id: check_v4
+      run: |
+        src_hash=$(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 pull --rebase upstream v4
+        git push upstream v4 --progress
+        git push origin v4 --progress
+        cd ../
+        rm -rf repo

.github/workflows/job_test.yml

@@ -0,0 +1,31 @@
+name: test
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches: [ master, v3, v4 ]
+  push:
+    branches: [ master, v3, v4 ]
+    paths-ignore:
+      - '.github/**'
+      - '.gitea/**'
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+    - name: checkout code
+      uses: actions/checkout@v4
+      with:
+        filter: 'blob:none'
+    - name: setup go
+      uses: actions/setup-go@v5
+      with:
+        cache-dependency-path: "**/*.sum"
+        go-version: 'stable'
+    - name: setup deps
+      run: go get -v ./...
+    - name: run test
+      env:
+        INTEGRATION_TESTS: yes
+      run: go test -mod readonly -v ./...

.github/workflows/job_tests.yml

@@ -0,0 +1,50 @@
+name: test
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches: [ master, v3, v4 ]
+  push:
+    branches: [ master, v3, v4 ]
+    paths-ignore:
+      - '.github/**'
+      - '.gitea/**'
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+    - name: checkout code
+      uses: actions/checkout@v4
+      with:
+        filter: 'blob:none'
+    - name: checkout tests
+      uses: actions/checkout@v4
+      with:
+        ref: master
+        filter: 'blob:none'
+        repository: unistack-org/micro-tests
+        path: micro-tests
+    - name: setup go
+      uses: actions/setup-go@v5
+      with:
+        cache-dependency-path: "**/*.sum"
+        go-version: 'stable'
+    - name: setup go work
+      env:
+        GOWORK: ${{ github.workspace }}/go.work
+      run: |
+        go work init
+        go work use .
+        go work use micro-tests
+    - name: setup deps
+      env:
+        GOWORK: ${{ github.workspace }}/go.work
+      run: go get -v ./...
+    - name: run tests
+      env:
+        INTEGRATION_TESTS: yes
+        GOWORK: ${{ github.workspace }}/go.work
+      run: |
+        cd micro-tests
+        go test -mod readonly -v ./... || true

.github/workflows/pr.yml

@@ -1,47 +0,0 @@
-name: prbuild
-on:
-  pull_request:
-    branches:
-    - master
-    - v3
-jobs:
-  test:
-    name: test
-    runs-on: ubuntu-latest
-    steps:
-    - name: setup
-      uses: actions/setup-go@v3
-      with:
-        go-version: 1.17
-    - name: checkout
-      uses: actions/checkout@v3
-    - name: cache
-      uses: actions/cache@v3
-      with:
-        path: ~/go/pkg/mod
-        key: ${{ runner.os }}-go-${{ hashFiles('**/go.sum') }}
-        restore-keys: ${{ runner.os }}-go-
-    - name: deps
-      run: go get -v -t -d ./...
-    - name: test
-      env:
-        INTEGRATION_TESTS: yes
-      run: go test -mod readonly -v ./...
-  lint:
-    name: lint
-    runs-on: ubuntu-latest
-    steps:
-      - name: checkout
-        uses: actions/checkout@v3
-      - name: lint
-        uses: golangci/golangci-lint-action@v3.4.0
-        continue-on-error: true
-        with:
-          # Required: the version of golangci-lint is required and must be specified without patch version: we always use the latest patch version.
-          version: v1.30
-          # Optional: working directory, useful for monorepos
-          # working-directory: somedir
-          # Optional: golangci-lint command line arguments.
-          # args: --issues-exit-code=0
-          # Optional: show only new issues if it's a pull request. The default value is `false`.
-          # only-new-issues: true

.gitignore

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

.golangci.yml

@@ -0,0 +1,5 @@
+run:
+  concurrency: 8
+  timeout: 5m
+  issues-exit-code: 1
+  tests: true

README.md

@@ -1,2 +1,2 @@
-# broker-kgo
-
+# micro-broker-kgo
+![Coverage](https://img.shields.io/badge/Coverage-64.8%25-yellow)

carrier.go

@@ -1,7 +1,12 @@
 package kgo
 
 import (
+	"net/http"
+	"slices"
+	"strings"
+
 	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v4/metadata"
 )
 
 // RecordCarrier injects and extracts traces from a kgo.Record.
@@ -51,3 +56,38 @@ func (c RecordCarrier) Keys() []string {
 	}
 	return out
 }
+
+func setHeaders(r *kgo.Record, md metadata.Metadata, exclude ...string) {
+	seen := make(map[string]struct{})
+
+loop:
+	for k, v := range md {
+		k = http.CanonicalHeaderKey(k)
+
+		if _, ok := seen[k]; ok {
+			continue loop
+		}
+
+		if slices.ContainsFunc(exclude, func(s string) bool {
+			return strings.EqualFold(s, k)
+		}) {
+			continue loop
+		}
+
+		for i := 0; i < len(r.Headers); i++ {
+			if strings.EqualFold(r.Headers[i].Key, k) {
+				// Key exist, update the value.
+				r.Headers[i].Value = []byte(strings.Join(v, ","))
+				continue loop
+			}
+		}
+
+		// Key does not exist, append new header.
+		r.Headers = append(r.Headers, kgo.RecordHeader{
+			Key:   k,
+			Value: []byte(strings.Join(v, ",")),
+		})
+
+		seen[k] = struct{}{}
+	}
+}

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)
+}

event.go

@@ -1,42 +0,0 @@
-package kgo
-
-import (
-	"sync"
-
-	"go.unistack.org/micro/v3/broker"
-)
-
-type event struct {
-	topic string
-	err   error
-	sync.RWMutex
-	msg *broker.Message
-	ack bool
-}
-
-func (p *event) Topic() string {
-	return p.topic
-}
-
-func (p *event) Message() *broker.Message {
-	return p.msg
-}
-
-func (p *event) Ack() error {
-	p.ack = true
-	return nil
-}
-
-func (p *event) Error() error {
-	return p.err
-}
-
-func (p *event) SetError(err error) {
-	p.err = err
-}
-
-var eventPool = sync.Pool{
-	New: func() interface{} {
-		return &event{msg: &broker.Message{}}
-	},
-}

go.mod

@@ -1,24 +1,27 @@
-module go.unistack.org/micro-broker-kgo/v3
+module go.unistack.org/micro-broker-kgo/v4
 
-go 1.21
-
-toolchain go1.22.1
+go 1.24.0
 
 require (
-	github.com/google/uuid v1.6.0
-	github.com/twmb/franz-go v1.16.1
-	github.com/twmb/franz-go/pkg/kadm v1.11.0
-	github.com/twmb/franz-go/pkg/kmsg v1.7.0
-	go.opentelemetry.io/otel v1.25.0
-	go.unistack.org/micro/v3 v3.10.59
+	github.com/stretchr/testify v1.11.1
+	github.com/twmb/franz-go v1.20.2
+	github.com/twmb/franz-go/pkg/kadm v1.17.1
+	github.com/twmb/franz-go/pkg/kfake v0.0.0-20250508175730-72e1646135e3
+	github.com/twmb/franz-go/pkg/kmsg v1.12.0
+	go.opentelemetry.io/otel v1.38.0
+	go.unistack.org/micro/v4 v4.1.24
 )
 
 require (
-	github.com/klauspost/compress v1.17.8 // indirect
-	github.com/pierrec/lz4/v4 v4.1.21 // indirect
-	golang.org/x/crypto v0.22.0 // indirect
-	golang.org/x/sys v0.19.0 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56 // indirect
-	google.golang.org/grpc v1.63.2 // indirect
-	google.golang.org/protobuf v1.33.0 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
+	github.com/google/uuid v1.6.0 // indirect
+	github.com/klauspost/compress v1.18.1 // indirect
+	github.com/matoous/go-nanoid v1.5.1 // indirect
+	github.com/pierrec/lz4/v4 v4.1.22 // 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 v0.43.0 // indirect
+	google.golang.org/protobuf v1.36.10 // indirect
+	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -1,40 +1,59 @@
-github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
-github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
-github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
+github.com/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=
+github.com/frankban/quicktest v1.14.6/go.mod h1:4ptaffx2x8+WTWXmUCuVU6aPUX1/Mz7zb5vbUoiM6w0=
+github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
+github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX3N/iU=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
-github.com/klauspost/compress v1.17.8 h1:YcnTYrq7MikUT7k0Yb5eceMmALQPYBW/Xltxn0NAMnU=
-github.com/klauspost/compress v1.17.8/go.mod h1:Di0epgTjJY877eYKx5yC51cX2A2Vl2ibi7bDH9ttBbw=
-github.com/pierrec/lz4/v4 v4.1.21 h1:yOVMLb6qSIDP67pl/5F7RepeKYu/VmTyEXvuMI5d9mQ=
-github.com/pierrec/lz4/v4 v4.1.21/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
-github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
-github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg=
-github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
-github.com/twmb/franz-go v1.16.1 h1:rpWc7fB9jd7TgmCyfxzenBI+QbgS8ZfJOUQE+tzPtbE=
-github.com/twmb/franz-go v1.16.1/go.mod h1:/pER254UPPGp/4WfGqRi+SIRGE50RSQzVubQp6+N4FA=
-github.com/twmb/franz-go/pkg/kadm v1.11.0 h1:FfeWJ0qadntFpAcQt8JzNXW4dijjytZNLrzJuzzzuxA=
-github.com/twmb/franz-go/pkg/kadm v1.11.0/go.mod h1:qrhkdH+SWS3ivmbqOgHbpgVHamhaKcjH0UM+uOp0M1A=
-github.com/twmb/franz-go/pkg/kmsg v1.7.0 h1:a457IbvezYfA5UkiBvyV3zj0Is3y1i8EJgqjJYoij2E=
-github.com/twmb/franz-go/pkg/kmsg v1.7.0/go.mod h1:se9Mjdt0Nwzc9lnjJ0HyDtLyBnaBDAd7pCje47OhSyw=
-go.opentelemetry.io/otel v1.25.0 h1:gldB5FfhRl7OJQbUHt/8s0a7cE8fbsPAtdpRaApKy4k=
-go.opentelemetry.io/otel v1.25.0/go.mod h1:Wa2ds5NOXEMkCmUou1WA7ZBfLTHWIsp034OVD7AO+Vg=
-go.unistack.org/micro/v3 v3.10.59 h1:eneYXJLgyu5MZpSvyI0K17CeXvgOoUCN5dWZaPV5lI4=
-go.unistack.org/micro/v3 v3.10.59/go.mod h1:erMgt3Bl7vQQ0e9UpQyR5NlLiZ9pKeEJ9+1tfYFaqUg=
-golang.org/x/crypto v0.22.0 h1:g1v0xeRhjcugydODzvb3mEM9SQ0HGp9s/nh3COQ/C30=
-golang.org/x/crypto v0.22.0/go.mod h1:vr6Su+7cTlO45qkww3VDJlzDn0ctJvRgYbC2NvXHt+M=
-golang.org/x/net v0.21.0 h1:AQyQV4dYCvJ7vGmJyKki9+PBdyvhkSd8EIx/qb0AYv4=
-golang.org/x/net v0.21.0/go.mod h1:bIjVDfnllIU7BJ2DNgfnXvpSvtn8VRwhlsaeUTyUS44=
-golang.org/x/sys v0.19.0 h1:q5f1RH2jigJ1MoAWp2KTp3gm5zAGFUTarQZ5U386+4o=
-golang.org/x/sys v0.19.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
-golang.org/x/text v0.14.0 h1:ScX5w1eTa3QqT8oi6+ziP7dTV1S2+ALU0bI+0zXKWiQ=
-golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56 h1:zviK8GX4VlMstrK3JkexM5UHjH1VOkRebH9y3jhSBGk=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20240412170617-26222e5d3d56/go.mod h1:WtryC6hu0hhx87FDGxWCDptyssuo68sk10vYjF+T9fY=
-google.golang.org/grpc v1.63.2 h1:MUeiw1B2maTVZthpU5xvASfTh3LDbxHd6IJ6QQVU+xM=
-google.golang.org/grpc v1.63.2/go.mod h1:WAX/8DgncnokcFUldAxq7GeB5DXHDbMF+lLvDomNkRA=
-google.golang.org/protobuf v1.33.0 h1:uNO2rsAINq/JlFpSdYEKIZ0uKD/R9cpdv0T+yoGwGmI=
-google.golang.org/protobuf v1.33.0/go.mod h1:c6P6GXX6sHbq/GpV6MGZEdwhWPcYBgnhAHhKbcUYpos=
+github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
+github.com/klauspost/compress v1.18.0/go.mod h1:2Pp+KzxcywXVXMr50+X0Q/Lsb43OQHYWRCY2AiWywWQ=
+github.com/klauspost/compress v1.18.1 h1:bcSGx7UbpBqMChDtsF28Lw6v/G94LPrrbMbdC3JH2co=
+github.com/klauspost/compress v1.18.1/go.mod h1:ZQFFVG+MdnR0P+l6wpXgIL4NTtwiKIdBnrBd8Nrxr+0=
+github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
+github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
+github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
+github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
+github.com/matoous/go-nanoid v1.5.1 h1:aCjdvTyO9LLnTIi0fgdXhOPPvOHjpXN6Ik9DaNjIct4=
+github.com/matoous/go-nanoid v1.5.1/go.mod h1:zyD2a71IubI24efhpvkJz+ZwfwagzgSO6UNiFsZKN7U=
+github.com/pierrec/lz4/v4 v4.1.22 h1:cKFw6uJDK+/gfw5BcDL0JL5aBsAFdsIT18eRtLj7VIU=
+github.com/pierrec/lz4/v4 v4.1.22/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
+github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 h1:Jamvg5psRIccs7FGNTlIRMkT8wgtp5eCXdBlqhYGL6U=
+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 v1.10.0 h1:h2x0u2shc1QuLHfxi+cTJvs30+ZAHOGRic8uyGTDWxY=
+github.com/spf13/cast v1.10.0/go.mod h1:jNfB8QC9IA6ZuY2ZjDp0KtFO2LZZlg4S/7bzP6qqeHo=
+github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
+github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
+github.com/twmb/franz-go v1.19.5 h1:W7+o8D0RsQsedqib71OVlLeZ0zI6CbFra7yTYhZTs5Y=
+github.com/twmb/franz-go v1.19.5/go.mod h1:4kFJ5tmbbl7asgwAGVuyG1ZMx0NNpYk7EqflvWfPCpM=
+github.com/twmb/franz-go v1.20.2 h1:CiwhyKZHW6vqSHJkh+RTxFAJkio0jBjM/JQhx/HZ72A=
+github.com/twmb/franz-go v1.20.2/go.mod h1:YCnepDd4gl6vdzG03I5Wa57RnCTIC6DVEyMpDX/J8UA=
+github.com/twmb/franz-go/pkg/kadm v1.16.1 h1:IEkrhTljgLHJ0/hT/InhXGjPdmWfFvxp7o/MR7vJ8cw=
+github.com/twmb/franz-go/pkg/kadm v1.16.1/go.mod h1:Ue/ye1cc9ipsQFg7udFbbGiFNzQMqiH73fGC2y0rwyc=
+github.com/twmb/franz-go/pkg/kadm v1.17.1 h1:Bt02Y/RLgnFO2NP2HVP1kd2TFtGRiJZx+fSArjZDtpw=
+github.com/twmb/franz-go/pkg/kadm v1.17.1/go.mod h1:s4duQmrDbloVW9QTMXhs6mViTepze7JLG43xwPcAeTg=
+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=
+github.com/twmb/franz-go/pkg/kmsg v1.12.0 h1:CbatD7ers1KzDNgJqPbKOq0Bz/WLBdsTH75wgzeVaPc=
+github.com/twmb/franz-go/pkg/kmsg v1.12.0/go.mod h1:+DPt4NC8RmI6hqb8G09+3giKObE6uD2Eya6CfqBpeJY=
+go.opentelemetry.io/otel v1.38.0 h1:RkfdswUDRimDg0m2Az18RKOsnI8UDzppJAtj01/Ymk8=
+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 v4.1.21 h1:F9PrbI1BhXSDS0FopwcO5wWrT2Xh38w9VhVZZBHjfg8=
+go.unistack.org/micro/v4 v4.1.21/go.mod h1:nlBXTbx0rQrSZX4HPp2m57PHmpuGPWUd0O+jpUIiPto=
+go.unistack.org/micro/v4 v4.1.24 h1:PbkSWJS3ssB5A0y0tOdOw6u9e2Mk4yombF4yR0Jshvo=
+go.unistack.org/micro/v4 v4.1.24/go.mod h1:nlBXTbx0rQrSZX4HPp2m57PHmpuGPWUd0O+jpUIiPto=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
+golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
+google.golang.org/protobuf v1.36.10 h1:AYd7cD/uASjIL6Q9LiTjz8JLcrh/88q5UObnmY3aOOE=
+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=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

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")
+}

kadmtest.go

@@ -1,65 +0,0 @@
-//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

@@ -5,29 +5,39 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"math/rand"
+	"math/rand/v2"
+	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
-	"github.com/google/uuid"
+	"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/v3/broker"
-	"go.unistack.org/micro/v3/logger"
-	"go.unistack.org/micro/v3/metadata"
-	"go.unistack.org/micro/v3/semconv"
-	"go.unistack.org/micro/v3/tracer"
-	mrand "go.unistack.org/micro/v3/util/rand"
+	"go.unistack.org/micro/v4/broker"
+	"go.unistack.org/micro/v4/codec"
+	"go.unistack.org/micro/v4/logger"
+	"go.unistack.org/micro/v4/metadata"
+	"go.unistack.org/micro/v4/options"
+	"go.unistack.org/micro/v4/semconv"
+	"go.unistack.org/micro/v4/tracer"
+	"go.unistack.org/micro/v4/util/id"
+	mrand "go.unistack.org/micro/v4/util/rand"
 )
 
 var _ broker.Broker = (*Broker)(nil)
 
+var messagePool = sync.Pool{
+	New: func() interface{} {
+		return &kgoMessage{}
+	},
+}
+
 var ErrLostMessage = errors.New("message not marked for offsets commit and will be lost in next iteration")
 
 var DefaultRetryBackoffFn = func() func(int) time.Duration {
 	var rngMu sync.Mutex
-	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
 	return func(fails int) time.Duration {
 		const (
 			min = 100 * time.Millisecond
@@ -43,7 +53,7 @@ var DefaultRetryBackoffFn = func() func(int) time.Duration {
 		backoff := min * time.Duration(1<<(fails-1))
 
 		rngMu.Lock()
-		jitter := 0.8 + 0.4*rng.Float64()
+		jitter := 0.8 + 0.4*rand.Float64()
 		rngMu.Unlock()
 
 		backoff = time.Duration(float64(backoff) * jitter)
@@ -56,13 +66,30 @@ var DefaultRetryBackoffFn = func() func(int) time.Duration {
 }()
 
 type Broker struct {
-	init      bool
+	funcPublish   broker.FuncPublish
+	funcSubscribe broker.FuncSubscribe
 	c             *kgo.Client
+	connected     *atomic.Uint32
+
 	kopts []kgo.Opt
-	connected bool
-	sync.RWMutex
+	subs  []*Subscriber
+
 	opts broker.Options
-	subs []*subscriber
+
+	mu   sync.RWMutex
+	init bool
+}
+
+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 {
@@ -73,7 +100,79 @@ 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
+}
+
+type kgoMessage struct {
+	c     codec.Codec
+	topic string
+	ctx   context.Context
+	body  []byte
+	hdr   metadata.Metadata
+	opts  broker.MessageOptions
+	ack   bool
+}
+
+func (m *kgoMessage) Ack() error {
+	m.ack = true
+	return nil
+}
+
+func (m *kgoMessage) Body() []byte {
+	return m.body
+}
+
+func (m *kgoMessage) Header() metadata.Metadata {
+	return m.hdr
+}
+
+func (m *kgoMessage) Context() context.Context {
+	return m.ctx
+}
+
+func (m *kgoMessage) Topic() string {
+	return ""
+}
+
+func (m *kgoMessage) Unmarshal(dst interface{}, opts ...codec.Option) error {
+	return m.c.Unmarshal(m.body, dst)
+}
+
+func (b *Broker) newCodec(ct string) (codec.Codec, error) {
+	if idx := strings.IndexRune(ct, ';'); idx >= 0 {
+		ct = ct[:idx]
+	}
+	b.mu.RLock()
+	c, ok := b.opts.Codecs[ct]
+	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.MessageOption) (broker.Message, error) {
+	options := broker.NewMessageOptions(opts...)
+	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)
+	}
+	if err != nil {
+		return nil, err
+	}
+
+	m.hdr.Set(metadata.HeaderContentType, m.opts.ContentType)
+
+	return m, nil
+}
+
+func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, *hookTracer, error) {
 	var c *kgo.Client
 	var err error
 
@@ -90,63 +189,133 @@ func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, err
 		}
 	}
 
+	var fatalOnError bool
+	if k.opts.Context != nil {
+		if v, ok := k.opts.Context.Value(fatalOnErrorKey{}).(bool); ok && v {
+			fatalOnError = v
+		}
+	}
+
+	htracer := &hookTracer{group: group, clientID: clientID, tracer: k.opts.Tracer}
 	opts = append(opts,
 		kgo.WithHooks(&hookMeter{meter: k.opts.Meter}),
-		kgo.WithHooks(&hookTracer{group: group, clientID: clientID, tracer: k.opts.Tracer}),
+		kgo.WithHooks(htracer),
+		kgo.WithHooks(&hookEvent{log: k.opts.Logger, fatalOnError: fatalOnError, connected: k.connected}),
 	)
 
 	select {
 	case <-ctx.Done():
 		if ctx.Err() != nil {
+			if sp != nil {
 				sp.SetStatus(tracer.SpanStatusError, ctx.Err().Error())
 			}
-		return nil, ctx.Err()
+		}
+		return nil, nil, ctx.Err()
 	default:
 		c, err = kgo.NewClient(opts...)
 		if err == nil {
 			err = c.Ping(ctx) // check connectivity to cluster
 		}
 		if err != nil {
+			if sp != nil {
 				sp.SetStatus(tracer.SpanStatusError, err.Error())
-			return nil, err
 			}
+			return nil, nil, err
+		}
+		k.connected.Store(1)
+		return c, htracer, nil
 	}
-	return c, nil
 }
 
 func (k *Broker) Connect(ctx context.Context) error {
-	k.RLock()
-	if k.connected {
-		k.RUnlock()
+	if k.connected.Load() == 1 {
 		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.mu.Lock()
 	k.c = c
-	k.connected = true
-	k.Unlock()
+	k.connected.Store(1)
+	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
 }
 
 func (k *Broker) Disconnect(ctx context.Context) error {
-	k.RLock()
-	if !k.connected {
-		k.RUnlock()
+	if k.connected.Load() == 0 {
 		return nil
 	}
-	k.RUnlock()
 
 	nctx := k.opts.Context
 	if ctx != nil {
@@ -156,8 +325,8 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 	ctx, span = k.opts.Tracer.Start(ctx, "Disconnect")
 	defer span.Finish()
 
-	k.Lock()
-	defer k.Unlock()
+	k.mu.Lock()
+	defer k.mu.Unlock()
 	select {
 	case <-nctx.Done():
 		return nctx.Err()
@@ -176,13 +345,13 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 		}
 	}
 
-	k.connected = false
+	k.connected.Store(0)
 	return nil
 }
 
 func (k *Broker) Init(opts ...broker.Option) error {
-	k.Lock()
-	defer k.Unlock()
+	k.mu.Lock()
+	defer k.mu.Unlock()
 
 	if len(opts) == 0 && k.init {
 		return nil
@@ -211,6 +380,18 @@ func (k *Broker) Init(opts ...broker.Option) error {
 		}
 	}
 
+	k.funcPublish = k.fnPublish
+	k.funcSubscribe = k.fnSubscribe
+
+	k.opts.Hooks.EachPrev(func(hook options.Hook) {
+		switch h := hook.(type) {
+		case broker.HookPublish:
+			k.funcPublish = h(k.funcPublish)
+		case broker.HookSubscribe:
+			k.funcSubscribe = h(k.funcSubscribe)
+		}
+	})
+
 	k.init = true
 
 	return nil
@@ -220,139 +401,178 @@ func (k *Broker) Options() broker.Options {
 	return k.opts
 }
 
-func (k *Broker) BatchPublish(ctx context.Context, msgs []*broker.Message, opts ...broker.PublishOption) error {
-	return k.publish(ctx, msgs, opts...)
+func (b *Broker) Publish(ctx context.Context, topic string, messages ...broker.Message) error {
+	return b.funcPublish(ctx, topic, messages...)
 }
 
-func (k *Broker) Publish(ctx context.Context, topic string, msg *broker.Message, opts ...broker.PublishOption) error {
-	msg.Header.Set(metadata.HeaderTopic, topic)
-	return k.publish(ctx, []*broker.Message{msg}, opts...)
+func (b *Broker) fnPublish(ctx context.Context, topic string, messages ...broker.Message) error {
+	return b.publish(ctx, topic, messages...)
 }
 
-func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...broker.PublishOption) error {
-	var span tracer.Span
-	ctx, span = k.opts.Tracer.Start(ctx, "Publish")
-	defer span.Finish()
+func (b *Broker) publish(ctx context.Context, topic string, messages ...broker.Message) error {
+	var records []*kgo.Record
 
-	k.Lock()
-	if !k.connected {
-		c, err := k.connect(ctx, k.kopts...)
-		if err != nil {
-			k.Unlock()
-			return err
-		}
-		k.c = c
-		k.connected = true
-	}
-	k.Unlock()
+	for _, msg := range messages {
+
+		rec := &kgo.Record{
+			Context: msg.Context(),
+			Topic:   topic,
+			Value:   msg.Body(),
+		}
 
-	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 rec.Context != nil {
+			if k, ok := rec.Context.Value(messageKey{}).([]byte); ok && k != nil {
+				rec.Key = k
 			}
-		if p, ok := options.Context.Value(publishPromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
+			if p, ok := rec.Context.Value(messagePromiseKey{}).(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)
-		msg.Header.Del(metadata.HeaderTopic)
-		k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Inc()
-		if options.BodyOnly || k.opts.Codec.String() == "noop" {
-			rec.Value = msg.Body
-			for k, v := range msg.Header {
-				rec.Headers = append(rec.Headers, kgo.RecordHeader{Key: k, Value: []byte(v)})
+		kmsg, ok := msg.(*kgoMessage)
+		if !ok {
+			continue
 		}
-		} else {
-			rec.Value, err = k.opts.Codec.Marshal(msg)
-			if err != nil {
-				return err
+		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
 			}
 		}
-		records = append(records, rec)
-	}
+
+		setHeaders(rec, msg.Header())
 
 		if promise != nil {
 			ts := time.Now()
-		for _, rec := range records {
-			k.c.Produce(ctx, rec, func(r *kgo.Record, err error) {
+			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)
-				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())
+				b.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Dec()
+				b.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", rec.Topic, "topic", rec.Topic).Update(te.Seconds())
+				b.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()
+					b.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()
+					b.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "success").Inc()
 				}
 				promise(r, err)
 			})
+			continue
+		} else {
+			records = append(records, rec)
 		}
-		return nil
 	}
+
+	if len(records) > 0 {
+		var errs []string
 		ts := time.Now()
-	results := k.c.ProduceSync(ctx, records...)
+		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 {
-		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()
+			b.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+			b.opts.Meter.Histogram(semconv.PublishMessageDurationSeconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+			b.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()
+				b.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()
+				b.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "success").Inc()
 			}
 		}
 
 		if len(errs) > 0 {
 			return fmt.Errorf("publish error: %s", strings.Join(errs, "\n"))
 		}
+	}
 
 	return nil
 }
 
-func (k *Broker) BatchSubscribe(ctx context.Context, topic string, handler broker.BatchHandler, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
-	return nil, 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 (b *Broker) Subscribe(ctx context.Context, topic string, handler interface{}, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
+	return b.funcSubscribe(ctx, topic, handler, opts...)
+}
+
+func (b *Broker) fnSubscribe(ctx context.Context, topic string, handler interface{}, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
+	if err := broker.IsValidHandler(handler); err != nil {
+		return nil, err
 	}
 
-func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Handler, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
 	options := broker.NewSubscribeOptions(opts...)
 
+	switch handler.(type) {
+	default:
+		return nil, broker.ErrInvalidHandler
+	case func(broker.Message) error:
+		break
+	case func([]broker.Message) error:
+		break
+	}
+
 	if options.Group == "" {
-		uid, err := uuid.NewRandom()
+		uid, err := id.New()
 		if err != nil {
 			return nil, err
 		}
-		options.Group = uid.String()
+		options.Group = uid
 	}
 
 	commitInterval := DefaultCommitInterval
-	if k.opts.Context != nil {
-		if v, ok := k.opts.Context.Value(commitIntervalKey{}).(time.Duration); ok && v > 0 {
+	if b.opts.Context != nil {
+		if v, ok := b.opts.Context.Value(commitIntervalKey{}).(time.Duration); ok && v > 0 {
 			commitInterval = v
 		}
 	}
 
-	sub := &subscriber{
+	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 {
+		if v, ok := options.Context.Value(fatalOnErrorKey{}).(bool); ok && v {
+			fatalOnError = v
+		}
+	}
+
+	sub := &Subscriber{
 		topic:        topic,
 		opts:         options,
 		handler:      handler,
-		kopts:     k.opts,
+		kopts:        b.opts,
 		consumers:    make(map[tp]*consumer),
 		done:         make(chan struct{}),
+		fatalOnError: fatalOnError,
+		connected:    b.connected,
+		messagePool:  messagePool,
 	}
 
-	kopts := append(k.kopts,
+	kopts := append(b.kopts,
 		kgo.ConsumerGroup(options.Group),
 		kgo.ConsumeTopics(topic),
 		kgo.ConsumeResetOffset(kgo.NewOffset().AtStart()),
@@ -360,7 +580,9 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		kgo.AutoCommitInterval(commitInterval),
 		kgo.OnPartitionsAssigned(sub.assigned),
 		kgo.OnPartitionsRevoked(sub.revoked),
+		kgo.StopProducerOnDataLossDetected(),
 		kgo.OnPartitionsLost(sub.lost),
+		kgo.AutoCommitCallback(sub.autocommit),
 		kgo.AutoCommitMarks(),
 	)
 
@@ -370,7 +592,7 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	c, err := k.connect(ctx, kopts...)
+	c, htracer, err := b.connect(ctx, kopts...)
 	if err != nil {
 		return nil, err
 	}
@@ -388,12 +610,14 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	}
 
 	sub.c = c
+	sub.htracer = htracer
 
 	go sub.poll(ctx)
 
-	k.Lock()
-	k.subs = append(k.subs, sub)
-	k.Unlock()
+	b.mu.Lock()
+	b.subs = append(b.subs, sub)
+	b.mu.Unlock()
+
 	return sub, nil
 }
 
@@ -414,12 +638,13 @@ func NewBroker(opts ...broker.Option) *Broker {
 		kgo.DialTimeout(3 * time.Second),
 		kgo.DisableIdempotentWrite(),
 		kgo.ProducerBatchCompression(kgo.NoCompression()),
-		kgo.WithLogger(&mlogger{l: options.Logger.Clone(logger.WithCallerSkipCount(options.Logger.Options().CallerSkipCount + 2)), ctx: options.Context}),
+		kgo.WithLogger(&mlogger{l: options.Logger.Clone(logger.WithAddCallerSkipCount(2)), ctx: options.Context}),
 		kgo.SeedBrokers(kaddrs...),
 		kgo.RetryBackoffFn(DefaultRetryBackoffFn),
 		kgo.BlockRebalanceOnPoll(),
 		kgo.Balancers(kgo.CooperativeStickyBalancer()),
 		kgo.FetchIsolationLevel(kgo.ReadUncommitted()),
+		kgo.UnknownTopicRetries(1),
 	}
 
 	if options.Context != nil {
@@ -429,6 +654,7 @@ func NewBroker(opts ...broker.Option) *Broker {
 	}
 
 	return &Broker{
+		connected: &atomic.Uint32{},
 		opts:      options,
 		kopts:     kopts,
 	}

kgo_test.go

@@ -2,39 +2,140 @@ package kgo_test
 
 import (
 	"context"
-	"fmt"
-	"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/v3"
-	"go.unistack.org/micro/v3/broker"
-	"go.unistack.org/micro/v3/logger"
-	"go.unistack.org/micro/v3/metadata"
+	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"
 )
 
 var (
-	msgcnt   = int64(12000000)
+	msgcnt   = int64(1200)
 	group    = "38"
-	prefill  = false
-	loglevel = logger.InfoLevel
+	prefill  = true
+	loglevel = logger.ErrorLevel
+	cluster  *kfake.Cluster
 )
 
-var bm = &broker.Message{
-	Header: map[string]string{"hkey": "hval", metadata.HeaderTopic: "test"},
-	Body:   []byte(`"body"`),
+func TestMain(m *testing.M) {
+	cluster = kfake.MustCluster(
+		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()
+
+	b := kgo.NewBroker(
+		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),
+			kg.AllowAutoTopicCreation(),
+		),
+	)
+
+	t.Logf("broker init")
+	if err := b.Init(); err != nil {
+		t.Fatal(err)
+	}
+
+	t.Logf("broker connect")
+	if err := b.Connect(ctx); err != nil {
+		t.Fatal(err)
+	}
+
+	defer func() {
+		t.Logf("broker disconnect")
+		if err := b.Disconnect(ctx); err != nil {
+			t.Fatal(err)
+		}
+	}()
+
+	t.Logf("broker health %v", b.Health())
+	msgs := make([]broker.Message, 0, msgcnt)
+	for i := int64(0); i < msgcnt; i++ {
+		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")
+			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(100 * time.Millisecond)
+		// t.Logf("ack")
+		return msg.Ack()
+	}
+
+	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)
+		}
+	}()
+
+	for {
+		t.Logf("health check")
+		if !b.Health() {
+			t.Logf("health works")
+			break
+		}
+		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)
@@ -46,27 +147,23 @@ 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), logger.WithCallerSkipCount(3)); err != nil {
+	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 {
 		t.Fatal(err)
 	}
@@ -81,25 +178,26 @@ func TestPubSub(t *testing.T) {
 		}
 	}()
 	if prefill {
-		msgs := make([]*broker.Message, 0, msgcnt)
+		msgs := make([]broker.Message, 0, msgcnt)
 		for i := int64(0); i < msgcnt; i++ {
-			msgs = append(msgs, bm)
+			m, _ := b.NewMessage(ctx, metadata.Pairs("hkey", "hval"), []byte(`test`))
+			msgs = append(msgs, m)
 		}
 
-		if err := b.BatchPublish(ctx, msgs); err != nil {
+		if err := b.Publish(ctx, "test.pubsub", msgs...); err != nil {
 			t.Fatal(err)
 		}
 		//	t.Skip()
 	}
 	done := make(chan bool, 1)
 	idx := int64(0)
-	fn := func(msg broker.Event) error {
+	fn := func(msg broker.Message) error {
 		atomic.AddInt64(&idx, 1)
 		// time.Sleep(200 * time.Millisecond)
 		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))
@@ -124,7 +222,7 @@ func TestPubSub(t *testing.T) {
 				if prc := atomic.LoadInt64(&idx); prc == msgcnt {
 					close(done)
 				} else {
-					fmt.Printf("processed %v\n", prc)
+					t.Logf("processed %v of %v\n", prc, msgcnt)
 				}
 			case <-ticker.C:
 				close(done)

logger.go

@@ -4,7 +4,7 @@ import (
 	"context"
 
 	"github.com/twmb/franz-go/pkg/kgo"
-	"go.unistack.org/micro/v3/logger"
+	"go.unistack.org/micro/v4/logger"
 )
 
 type mlogger struct {
@@ -28,11 +28,8 @@ func (l *mlogger) Log(lvl kgo.LogLevel, msg string, args ...interface{}) {
 	default:
 		return
 	}
-	if len(args) > 0 {
-		l.l.Log(l.ctx, mlvl, append([]interface{}{msg}, args...)...)
-	} else {
-		l.l.Log(l.ctx, mlvl, msg)
-	}
+
+	l.l.Log(l.ctx, mlvl, msg, args...)
 }
 
 func (l *mlogger) Level() kgo.LogLevel {

meter.go

@@ -6,7 +6,7 @@ import (
 	"time"
 
 	"github.com/twmb/franz-go/pkg/kgo"
-	"go.unistack.org/micro/v3/meter"
+	"go.unistack.org/micro/v4/meter"
 )
 
 type hookMeter struct {
@@ -61,7 +61,7 @@ const (
 	labelTopic   = "topic"
 )
 
-func (m *hookMeter) OnGroupManageError(err error) {
+func (m *hookMeter) OnGroupManageError(_ error) {
 	m.meter.Counter(metricBrokerGroupErrors).Inc()
 }
 

options.go

@@ -5,8 +5,7 @@ import (
 	"time"
 
 	"github.com/twmb/franz-go/pkg/kgo"
-	"go.unistack.org/micro/v3/broker"
-	"go.unistack.org/micro/v3/client"
+	"go.unistack.org/micro/v4/broker"
 )
 
 var (
@@ -15,7 +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
@@ -28,16 +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)
-func PublishKey(key []byte) broker.PublishOption {
-	return broker.SetPublishOption(publishKey{}, key)
-}
-
-// ClientPublishKey set the kafka message key (client option)
-func ClientPublishKey(key []byte) client.PublishOption {
-	return client.SetPublishOption(publishKey{}, key)
+// MessageKey set the kafka message key (broker option)
+func MessageKey(key []byte) broker.MessageOption {
+	return broker.SetMessageOption(messageKey{}, key)
 }
 
 type optionsKey struct{}
@@ -72,6 +66,12 @@ func SubscribeOptions(opts ...kgo.Opt) broker.SubscribeOption {
 	}
 }
 
+type fatalOnErrorKey struct{}
+
+func FatalOnError(b bool) broker.Option {
+	return broker.SetOption(fatalOnErrorKey{}, b)
+}
+
 type clientIDKey struct{}
 
 func ClientID(id string) broker.Option {
@@ -98,14 +98,28 @@ 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)
+// SubscribeMaxInFlight max queued messages
+func SubscribeFatalOnError(b bool) broker.SubscribeOption {
+	return broker.SetSubscribeOption(fatalOnErrorKey{}, b)
 }
 
-// ClientPublishKey set the kafka message key (client option)
-func ClientPublishPromise(fn func(*kgo.Record, error)) client.PublishOption {
-	return client.SetPublishOption(publishPromiseKey{}, fn)
+type messagePromiseKey struct{}
+
+// MessagePromise set the kafka promise func for Produce
+func MessagePromise(fn func(*kgo.Record, error)) broker.MessageOption {
+	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

@@ -2,16 +2,19 @@ package kgo
 
 import (
 	"context"
+	"fmt"
 	"strconv"
 	"sync"
+	"sync/atomic"
 	"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"
+	"github.com/twmb/franz-go/pkg/kmsg"
+	"go.unistack.org/micro/v4/broker"
+	"go.unistack.org/micro/v4/logger"
+	"go.unistack.org/micro/v4/metadata"
+	"go.unistack.org/micro/v4/semconv"
+	"go.unistack.org/micro/v4/tracer"
 )
 
 type tp struct {
@@ -20,45 +23,53 @@ type tp struct {
 }
 
 type consumer struct {
-	c         *kgo.Client
 	topic       string
-	partition int32
-	opts      broker.SubscribeOptions
-	kopts     broker.Options
-	handler   broker.Handler
+	c           *kgo.Client
+	htracer     *hookTracer
 	quit        chan struct{}
 	done        chan struct{}
 	recs        chan kgo.FetchTopicPartition
-}
-
-type subscriber struct {
-	c         *kgo.Client
-	topic     string
-	opts      broker.SubscribeOptions
 	kopts       broker.Options
-	handler   broker.Handler
-	closed    bool
-	done      chan struct{}
-	consumers map[tp]*consumer
-	sync.RWMutex
+	partition   int32
+	opts        broker.SubscribeOptions
+	handler     interface{}
+	connected   *atomic.Uint32
+	messagePool bool
 }
 
-func (s *subscriber) Options() broker.SubscribeOptions {
+type Subscriber struct {
+	consumers    map[tp]*consumer
+	c            *kgo.Client
+	htracer      *hookTracer
+	topic        string
+	messagePool  bool
+	handler      interface{}
+	done         chan struct{}
+	kopts        broker.Options
+	opts         broker.SubscribeOptions
+	connected    *atomic.Uint32
+	mu           sync.RWMutex
+	closed       bool
+	fatalOnError bool
+}
+
+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():
-	//	return ctx.Err()
-	default:
+
 	s.c.PauseFetchTopics(s.topic)
 	s.c.CloseAllowingRebalance()
 	kc := make(map[string][]int32)
@@ -69,52 +80,19 @@ func (s *subscriber) Unsubscribe(ctx context.Context) error {
 	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
-				}
-
-				for tp := range s.consumers {
-					if v, ok := lmap[tp.p]; ok {
-						s.kopts.Meter.Counter(semconv.BrokerGroupLag, "topic", s.topic, "group", s.opts.Group, "partition", strconv.Itoa(int(tp.p)), "lag", strconv.Itoa(int(v.Lag)))
-					}
-				}
-
-			}
-		}
-	}()
-
 	for {
 		select {
 		case <-ctx.Done():
@@ -129,164 +107,182 @@ 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)
+				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) {
-				tp := tp{p.Topic, p.Partition}
-				s.consumers[tp].recs <- p
+				tps := tp{p.Topic, p.Partition}
+				s.mu.Lock()
+				c := s.consumers[tps]
+				s.mu.Unlock()
+				if c != nil {
+					c.recs <- p
+				}
 			})
 			s.c.AllowRebalance()
 		}
 	}
 }
 
-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()
 
 	for topic, partitions := range lost {
 		for _, partition := range partitions {
-			tp := tp{topic, partition}
-			pc := s.consumers[tp]
-			delete(s.consumers, tp)
+			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)
-			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) {
-	s.kopts.Logger.Debugf(ctx, "[kgo] lost %#+v", lost)
-	s.killConsumers(ctx, lost)
+func (s *Subscriber) autocommit(_ *kgo.Client, _ *kmsg.OffsetCommitRequest, _ *kmsg.OffsetCommitResponse, err error) {
+	if err != nil {
+		//		s.connected.Store(0)
+		if s.fatalOnError {
+			s.kopts.Logger.Fatal(context.TODO(), "kgo.AutoCommitCallback error", err)
+		}
+	}
 }
 
-func (s *subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[string][]int32) {
-	s.kopts.Logger.Debugf(ctx, "[kgo] revoked %#+v", revoked)
+func (s *Subscriber) lost(ctx context.Context, _ *kgo.Client, lost map[string][]int32) {
+	if s.kopts.Logger.V(logger.ErrorLevel) {
+		s.kopts.Logger.Error(ctx, fmt.Sprintf("[kgo] lost %#+v", lost))
+	}
+	s.killConsumers(ctx, lost)
+	// s.connected.Store(0)
+}
+
+func (s *Subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[string][]int32) {
+	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)
+		// s.connected.Store(0)
 	}
 }
 
-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{}),
 				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.mu.Lock()
 			s.consumers[tp{topic, partition}] = pc
+			s.mu.Unlock()
 			go pc.consume()
 		}
 	}
 }
 
 func (pc *consumer) consume() {
-	defer close(pc.done)
-	pc.kopts.Logger.Debugf(pc.kopts.Context, "starting, topic %s partition %d", pc.topic, pc.partition)
-	defer pc.kopts.Logger.Debugf(pc.kopts.Context, "killing, topic %s partition %d", pc.topic, pc.partition)
+	var err error
 
-	eh := pc.kopts.ErrorHandler
-	if pc.opts.ErrorHandler != nil {
-		eh = pc.opts.ErrorHandler
+	defer close(pc.done)
+	if pc.kopts.Logger.V(logger.DebugLevel) {
+		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))
 	}
 
+	var pm *kgoMessage
+
 	for {
 		select {
 		case <-pc.quit:
 			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))
+
+				if pc.messagePool {
+					pm = messagePool.Get().(*kgoMessage)
+				} else {
+					pm = &kgoMessage{}
+				}
+				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 {
-					p.msg.Header.Set(hdr.Key, string(hdr.Value))
+					pm.hdr.Set(hdr.Key, string(hdr.Value))
 				}
-				if pc.kopts.Codec.String() == "noop" {
-					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 {
-						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)
+				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(pm)
+				case func([]broker.Message) error:
+					err = h([]broker.Message{pm})
+				}
+
 				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
-							if p.ack {
+				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 {
+					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 := pm.ack
+				if pc.messagePool {
+					messagePool.Put(p)
+				}
+				if ack {
 					pc.c.MarkCommitRecords(record)
 				} else {
-								eventPool.Put(p)
-								pc.kopts.Logger.Fatalf(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+					if sp != nil {
+						sp.Finish()
+					}
+					//					pc.connected.Store(0)
+					pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] message not commited")
 					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.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.Fatalf(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
-						return
-					}
-				}
-				err := pc.handler(p)
-				if err == nil {
-					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "success").Inc()
-				} else {
-					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 {
-						_ = eh(p)
-					} else {
-						if pc.kopts.Logger.V(logger.ErrorLevel) {
-							pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: subscriber error: %v", err)
-						}
-					}
-				}
-				te := time.Since(ts)
-				pc.kopts.Meter.Summary(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?")
-					return
+				if sp != nil {
+					sp.Finish()
 				}
 			}
 		}

tracer.go

@@ -2,56 +2,29 @@ package kgo
 
 import (
 	"context"
-	"net"
-	"time"
 	"unicode/utf8"
 
 	"github.com/twmb/franz-go/pkg/kgo"
 	semconv "go.opentelemetry.io/otel/semconv/v1.18.0"
-	"go.unistack.org/micro/v3/tracer"
+	"go.unistack.org/micro/v4/metadata"
+	"go.unistack.org/micro/v4/tracer"
 )
 
 type hookTracer struct {
+	tracer   tracer.Tracer
 	clientID string
 	group    string
-	tracer   tracer.Tracer
 }
 
+var messagingSystem = semconv.MessagingSystemKey.String("kafka")
+
 var (
-	_ kgo.HookBrokerConnect       = &hookTracer{}
-	_ kgo.HookBrokerDisconnect    = &hookTracer{}
-	_ kgo.HookBrokerRead          = &hookTracer{}
-	_ kgo.HookBrokerThrottle      = &hookTracer{}
-	_ kgo.HookBrokerWrite         = &hookTracer{}
-	_ kgo.HookFetchBatchRead      = &hookTracer{}
-	_ kgo.HookProduceBatchWritten = &hookTracer{}
-	_ kgo.HookGroupManageError    = &hookTracer{}
+	_ kgo.HookProduceRecordBuffered   = (*hookTracer)(nil)
+	_ kgo.HookProduceRecordUnbuffered = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordBuffered     = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordUnbuffered   = (*hookTracer)(nil)
 )
 
-func (m *hookTracer) OnGroupManageError(err error) {
-}
-
-func (m *hookTracer) OnBrokerConnect(meta kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
-}
-
-func (m *hookTracer) OnBrokerDisconnect(meta kgo.BrokerMetadata, _ net.Conn) {
-}
-
-func (m *hookTracer) OnBrokerWrite(meta kgo.BrokerMetadata, _ int16, bytesWritten int, writeWait, timeToWrite time.Duration, err error) {
-}
-
-func (m *hookTracer) OnBrokerRead(meta kgo.BrokerMetadata, _ int16, bytesRead int, readWait, timeToRead time.Duration, err error) {
-}
-
-func (m *hookTracer) OnBrokerThrottle(meta kgo.BrokerMetadata, throttleInterval time.Duration, _ bool) {
-}
-
-func (m *hookTracer) OnProduceBatchWritten(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.ProduceBatchMetrics) {
-}
-
-func (m *hookTracer) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ int32, kmetrics kgo.FetchBatchMetrics) {
-}
-
 // OnProduceRecordBuffered starts a new span for the "publish" operation on a
 // buffered record.
 //
@@ -59,9 +32,12 @@ func (m *hookTracer) OnFetchBatchRead(meta kgo.BrokerMetadata, topic string, _ i
 // 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{}{
-		semconv.MessagingSystemKey.String("kafka"),
+		messagingSystem,
 		semconv.MessagingDestinationKindTopic,
 		semconv.MessagingDestinationName(r.Topic),
 		semconv.MessagingOperationPublish,
@@ -74,12 +50,28 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 		tracer.WithSpanLabels(attrs...),
 		tracer.WithSpanKind(tracer.SpanKindProducer),
 	}
-	// Start the "publish" span.
-	ctx, _ := m.tracer.Start(r.Context, r.Topic+" publish", opts...)
-	// Inject the span context into the record.
-	// t.propagators.Inject(ctx, NewRecordCarrier(r))
-	// Update the record context.
-	r.Context = ctx
+
+	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, metadata.HeaderContentType)
 }
 
 // OnProduceRecordUnbuffered continues and ends the "publish" span for an
@@ -88,17 +80,18 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 // It sets attributes with values unset when producing and records any error
 // that occurred during the publish operation.
 func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
-	span, ok := tracer.SpanFromContext(r.Context)
-	if !ok {
+	if !m.tracer.Enabled() {
 		return
 	}
-	defer span.Finish()
+	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
@@ -109,9 +102,12 @@ 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{}{
-		semconv.MessagingSystemKey.String("kafka"),
+		messagingSystem,
 		semconv.MessagingSourceKindTopic,
 		semconv.MessagingSourceName(r.Topic),
 		semconv.MessagingOperationReceive,
@@ -132,20 +128,33 @@ func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
 	if r.Context == nil {
 		r.Context = context.Background()
 	}
-	// Extract the span context from the record.
-	// ctx := t.propagators.Extract(r.Context, NewRecordCarrier(r))
-	// Start the "receive" span.
-	newCtx, _ := m.tracer.Start(r.Context, r.Topic+" receive", opts...)
-	// Update the record context.
-	r.Context = newCtx
+	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, metadata.HeaderContentType)
 }
 
 // OnFetchRecordUnbuffered continues and ends the "receive" span for an
 // unbuffered record.
 func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
-	if span, ok := tracer.SpanFromContext(r.Context); ok {
-		defer span.Finish()
+	if !m.tracer.Enabled() {
+		return
 	}
+	span, _ := tracer.SpanFromContext(r.Context)
+	span.Finish()
 }
 
 // WithProcessSpan starts a new span for the "process" operation on a consumer
@@ -158,9 +167,16 @@ 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{}{
-		semconv.MessagingSystemKey.String("kafka"),
+		messagingSystem,
 		semconv.MessagingSourceKindTopic,
 		semconv.MessagingSourceName(r.Topic),
 		semconv.MessagingOperationProcess,
@@ -182,8 +198,16 @@ func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Spa
 	if r.Context == nil {
 		r.Context = context.Background()
 	}
+	md, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		md = metadata.New(len(r.Headers))
+	}
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
 	// Start a new span using the provided context and options.
-	return m.tracer.Start(r.Context, r.Topic+" process", opts...)
+	return m.tracer.Start(r.Context, "sdk.broker", opts...)
 }
 
 func maybeKeyAttr(attrs []interface{}, r *kgo.Record) []interface{} {