Merge pull request 'added check span!=nil && conn close' (#154) from devstigneev/micro-broker-kgo:v3 into v3

Reviewed-on: #154

.gitea/workflows/job_lint.yml

@@ -0,0 +1,29 @@
+name: lint
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches:
+    - master
+    - v3
+    - v4
+
+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: https://github.com/golangci/golangci-lint-action@v6
+      with:
+        version: 'latest'

.gitea/workflows/job_test.yml

@@ -0,0 +1,34 @@
+name: test
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches:
+    - master
+    - v3
+    - v4
+  push:
+    branches:
+    - master
+    - v3
+    - v4
+
+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 ./...

.gitea/workflows/job_tests.yml

@@ -0,0 +1,53 @@
+name: test
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize]
+    branches:
+    - master
+    - v3
+    - v4
+  push:
+    branches:
+    - master
+    - v3
+    - v4
+
+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: /workspace/${{ github.repository_owner }}/go.work
+      run: |
+        go work init
+        go work use .
+        go work use micro-tests        
+    - name: setup deps
+      env:
+        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+      run: go get -v ./...
+    - name: run tests
+      env:
+        INTEGRATION_TESTS: yes
+        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+      run: |
+        cd micro-tests
+        go test -mod readonly -v ./... || true        

.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/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 +1,21 @@
+run:
+  concurrency: 8
+  deadline: 5m
+  issues-exit-code: 1
+  tests: true
+
+linters:
+  enable:
+    - staticcheck
+    - unused
+    - gosimple
+    - govet
+    - prealloc
+    - unconvert
+    - nakedret
+
+linters-settings:
+  govet:
+    check-all: true
+    enable:
+      - fieldalignment

README.md

@@ -1,9 +1,9 @@
-# micro-broker-kgo
-yet another micro kafka broker alternative
+# broker-kgo
 
 TODO:
-* dont always append options from context on Init and New
-* add SubscriberOptions(...kgo.Opt)
-* add ServerSubscribeOptions(...kgo.Opt)
-* check PublisherOptions(...kgo.Opt)
-* check ClientPublisherOptions(...kgo.Opt)
+
+1) экспортируем текущий оффсет для каждой партиции в группе
+2) экспортируем лаг для группы
+3) мониторим
+   1) если есть лаг больше нуля
+   2) если дельта оффсета за нужное нам время не 

broker.go

@@ -0,0 +1,71 @@
+package kgo
+
+import (
+	"context"
+	"net"
+	"sync/atomic"
+	"time"
+
+	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v3/logger"
+)
+
+type hookEvent struct {
+	log          logger.Logger
+	connected    *atomic.Uint32
+	fatalOnError bool
+}
+
+var (
+	_ kgo.HookBrokerConnect           = &hookEvent{}
+	_ kgo.HookBrokerDisconnect        = &hookEvent{}
+	_ kgo.HookBrokerRead              = &hookEvent{}
+	_ kgo.HookBrokerWrite             = &hookEvent{}
+	_ kgo.HookGroupManageError        = &hookEvent{}
+	_ kgo.HookProduceRecordUnbuffered = &hookEvent{}
+)
+
+func (m *hookEvent) OnGroupManageError(err error) {
+	if err != nil {
+		// m.connected.Store(0)
+		// if m.fatalOnError {
+		m.log.Error(context.TODO(), "kgo.OnGroupManageError", err)
+		//}
+	}
+}
+
+func (m *hookEvent) OnBrokerConnect(_ kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
+	if err != nil {
+		// m.connected.Store(0)
+		// if m.fatalOnError {
+		m.log.Error(context.TODO(), "kgo.OnBrokerConnect", err)
+		//}
+	}
+}
+
+func (m *hookEvent) OnBrokerDisconnect(_ kgo.BrokerMetadata, _ net.Conn) {
+	// m.connected.Store(0)
+}
+
+func (m *hookEvent) OnBrokerWrite(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
+	if err != nil {
+		// m.connected.Store(0)
+		// if m.fatalOnError {
+		m.log.Error(context.TODO(), "kgo.OnBrokerWrite", err)
+		//}
+	}
+}
+
+func (m *hookEvent) OnBrokerRead(_ kgo.BrokerMetadata, _ int16, _ int, _ time.Duration, _ time.Duration, err error) {
+	if err != nil {
+		// m.connected.Store(0)
+		m.log.Error(context.TODO(), "kgo.OnBrokerRead", err)
+	}
+}
+
+func (m *hookEvent) OnProduceRecordUnbuffered(_ *kgo.Record, err error) {
+	if err != nil {
+		// m.connected.Store(0)
+		m.log.Error(context.TODO(), "kgo.OnProduceRecordUnbuffered", err)
+	}
+}

carrier.go

@@ -0,0 +1,93 @@
+package kgo
+
+import (
+	"net/http"
+	"slices"
+	"strings"
+
+	"github.com/twmb/franz-go/pkg/kgo"
+	"go.unistack.org/micro/v3/metadata"
+)
+
+// RecordCarrier injects and extracts traces from a kgo.Record.
+//
+// This type exists to satisfy the otel/propagation.TextMapCarrier interface.
+type RecordCarrier struct {
+	record *kgo.Record
+}
+
+// NewRecordCarrier creates a new RecordCarrier.
+func NewRecordCarrier(record *kgo.Record) RecordCarrier {
+	return RecordCarrier{record: record}
+}
+
+// Get retrieves a single value for a given key if it exists.
+func (c RecordCarrier) Get(key string) string {
+	for _, h := range c.record.Headers {
+		if h.Key == key {
+			return string(h.Value)
+		}
+	}
+	return ""
+}
+
+// Set sets a header.
+func (c RecordCarrier) Set(key, val string) {
+	// Check if key already exists.
+	for i, h := range c.record.Headers {
+		if h.Key == key {
+			// Key exist, update the value.
+			c.record.Headers[i].Value = []byte(val)
+			return
+		}
+	}
+	// Key does not exist, append new header.
+	c.record.Headers = append(c.record.Headers, kgo.RecordHeader{
+		Key:   key,
+		Value: []byte(val),
+	})
+}
+
+// Keys returns a slice of all key identifiers in the carrier.
+func (c RecordCarrier) Keys() []string {
+	out := make([]string, len(c.record.Headers))
+	for i, h := range c.record.Headers {
+		out[i] = h.Key
+	}
+	return out
+}
+
+func setHeaders(r *kgo.Record, md metadata.Metadata, 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(v)
+				continue loop
+			}
+		}
+
+		// Key does not exist, append new header.
+		r.Headers = append(r.Headers, kgo.RecordHeader{
+			Key:   k,
+			Value: []byte(v),
+		})
+
+		seen[k] = struct{}{}
+	}
+}

event.go

@@ -1,19 +1,27 @@
 package kgo
 
 import (
+	"context"
 	"sync"
 
 	"go.unistack.org/micro/v3/broker"
 )
 
 type event struct {
-	topic string
-	err   error
-	sync.RWMutex
 	msg *broker.Message
+	err error
+	ctx context.Context
+
+	topic string
+
+	sync.RWMutex
 	ack bool
 }
 
+func (p *event) Context() context.Context {
+	return p.ctx
+}
+
 func (p *event) Topic() string {
 	return p.topic
 }

go.mod

@@ -1,14 +1,27 @@
 module go.unistack.org/micro-broker-kgo/v3
 
-go 1.17
+go 1.23.0
 
 require (
-	github.com/twmb/franz-go v1.11.5
-	github.com/twmb/franz-go/pkg/kmsg v1.3.0
-	go.unistack.org/micro/v3 v3.10.14
+	github.com/google/uuid v1.6.0
+	github.com/twmb/franz-go v1.18.1
+	github.com/twmb/franz-go/pkg/kadm v1.15.0
+	github.com/twmb/franz-go/pkg/kmsg v1.9.0
+	go.opentelemetry.io/otel v1.34.0
+	go.unistack.org/micro/v3 v3.11.44
 )
 
 require (
-	github.com/klauspost/compress v1.15.9 // indirect
-	github.com/pierrec/lz4/v4 v4.1.15 // indirect
+	github.com/ash3in/uuidv8 v1.2.0 // indirect
+	github.com/klauspost/compress v1.18.0 // indirect
+	github.com/kr/pretty v0.3.1 // indirect
+	github.com/matoous/go-nanoid v1.5.1 // indirect
+	github.com/pierrec/lz4/v4 v4.1.22 // indirect
+	go.unistack.org/micro-proto/v3 v3.4.1 // indirect
+	golang.org/x/crypto v0.35.0 // indirect
+	golang.org/x/sys v0.30.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20250224174004-546df14abb99 // indirect
+	google.golang.org/grpc v1.70.0 // indirect
+	google.golang.org/protobuf v1.36.5 // indirect
+	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -1,24 +1,62 @@
-github.com/imdario/mergo v0.3.13/go.mod h1:4lJ1jqUDcsbIECGy0RUJAXNIhg+6ocWgb1ALK2O4oXg=
-github.com/klauspost/compress v1.15.9 h1:wKRjX6JRtDdrE9qwa4b/Cip7ACOshUI4smpCQanqjSY=
-github.com/klauspost/compress v1.15.9/go.mod h1:PhcZ0MbTNciWF3rruxRgKxI5NkcHHrHUDtV4Yw2GlzU=
-github.com/patrickmn/go-cache v2.1.0+incompatible/go.mod h1:3Qf8kWWT7OJRJbdiICTKqZju1ZixQ/KpMGzzAfe6+WQ=
-github.com/pierrec/lz4/v4 v4.1.15 h1:MO0/ucJhngq7299dKLwIMtgTfbkoSPF6AoMYDd8Q4q0=
-github.com/pierrec/lz4/v4 v4.1.15/go.mod h1:gZWDp/Ze/IJXGXf23ltt2EXimqmTUXEy0GFuRQyBid4=
-github.com/silas/dag v0.0.0-20211117232152-9d50aa809f35/go.mod h1:7RTUFBdIRC9nZ7/3RyRNH1bdqIShrDejd1YbLwgPS+I=
-github.com/twmb/franz-go v1.11.5 h1:TTv5lVJd+87XkmP9dWN9Jgpf7IUUr7a7jee+byR8LBE=
-github.com/twmb/franz-go v1.11.5/go.mod h1:FvaHNlpT6woVYIl6LAuIeL7yHol1Fp6Gv2Dn21AvH78=
-github.com/twmb/franz-go/pkg/kmsg v1.3.0 h1:ouBETB7nTqRxiO5E8/pySoFZtVEW2VWw55z3/bsUzTw=
-github.com/twmb/franz-go/pkg/kmsg v1.3.0/go.mod h1:SxG/xJKhgPu25SamAq0rrucfp7lbzCpEXOC+vH/ELrY=
-go.unistack.org/micro/v3 v3.10.14 h1:7fgLpwGlCN67twhwtngJDEQvrMkUBDSA5vzZqxIDqNE=
-go.unistack.org/micro/v3 v3.10.14/go.mod h1:uMAc0U/x7dmtICCrblGf0ZLgYegu3VwQAquu+OFCw1Q=
-golang.org/x/crypto v0.0.0-20220817201139-bc19a97f63c8/go.mod h1:IxCIyHEi3zRg3s0A5j5BB6A9Jmi73HwBIUl50j+osU4=
-golang.org/x/net v0.0.0-20211112202133-69e39bad7dc2/go.mod h1:9nx3DQGgdP8bBQD5qxJ1jj9UTztislL4KSBs9R2vV5Y=
-golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20210423082822-04245dca01da/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
-golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
+github.com/DATA-DOG/go-sqlmock v1.5.2 h1:OcvFkGmslmlZibjAjaHm3L//6LiuBgolP7OputlJIzU=
+github.com/DATA-DOG/go-sqlmock v1.5.2/go.mod h1:88MAG/4G7SMwSE3CeA0ZKzrT5CiOU3OJ+JlNzwDqpNU=
+github.com/ash3in/uuidv8 v1.2.0 h1:2oogGdtCPwaVtyvPPGin4TfZLtOGE5F+W++E880G6SI=
+github.com/ash3in/uuidv8 v1.2.0/go.mod h1:BnU0wJBxnzdEKmVg4xckBkD+VZuecTFTUP3M0dWgyY4=
+github.com/creack/pty v1.1.9/go.mod h1:oKZEueFk5CKHvIhNR5MUki03XCEU+Q6VDXinZuGJ33E=
+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/golang/protobuf v1.5.4 h1:i7eJL8qZTpSEXOPTxNKhASYpMn+8e5Q6AdndVa1dWek=
+github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6rSs7xps=
+github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
+github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
+github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
+github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
+github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
+github.com/klauspost/compress v1.18.0/go.mod h1:2Pp+KzxcywXVXMr50+X0Q/Lsb43OQHYWRCY2AiWywWQ=
+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/pkg/diff v0.0.0-20210226163009-20ebb0f2a09e/go.mod h1:pJLUxLENpZxwdsKMEsNbx1VGcRFpLqf3715MtcvvzbA=
+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 v1.9.0/go.mod h1:WtVeX8xhTBvf0smdhujwtBcq4Qrzq/fJaraNFVN+nFs=
+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/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
+github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
+github.com/twmb/franz-go v1.18.1 h1:D75xxCDyvTqBSiImFx2lkPduE39jz1vaD7+FNc+vMkc=
+github.com/twmb/franz-go v1.18.1/go.mod h1:Uzo77TarcLTUZeLuGq+9lNpSkfZI+JErv7YJhlDjs9M=
+github.com/twmb/franz-go/pkg/kadm v1.15.0 h1:Yo3NAPfcsx3Gg9/hdhq4vmwO77TqRRkvpUcGWzjworc=
+github.com/twmb/franz-go/pkg/kadm v1.15.0/go.mod h1:MUdcUtnf9ph4SFBLLA/XxE29rvLhWYLM9Ygb8dfSCvw=
+github.com/twmb/franz-go/pkg/kmsg v1.9.0 h1:JojYUph2TKAau6SBtErXpXGC7E3gg4vGZMv9xFU/B6M=
+github.com/twmb/franz-go/pkg/kmsg v1.9.0/go.mod h1:CMbfazviCyY6HM0SXuG5t9vOwYDHRCSrJJyBAe5paqg=
+go.opentelemetry.io/otel v1.34.0 h1:zRLXxLCgL1WyKsPVrgbSdMN4c0FMkDAskSTQP+0hdUY=
+go.opentelemetry.io/otel v1.34.0/go.mod h1:OWFPOQ+h4G8xpyjgqo4SxJYdDQ/qmRH+wivy7zzx9oI=
+go.unistack.org/micro-proto/v3 v3.4.1 h1:UTjLSRz2YZuaHk9iSlVqqsA50JQNAEK2ZFboGqtEa9Q=
+go.unistack.org/micro-proto/v3 v3.4.1/go.mod h1:okx/cnOhzuCX0ggl/vToatbCupi0O44diiiLLsZ93Zo=
+go.unistack.org/micro/v3 v3.11.44 h1:A+T8zVcL2vlL66kn/Y4rqhtBybLO829wFEYZJYorDOU=
+go.unistack.org/micro/v3 v3.11.44/go.mod h1:13EFW2ps3BN9mpYbp9K0oQu/VDjEN6LJ4wwdom7hcXQ=
+golang.org/x/crypto v0.35.0 h1:b15kiHdrGCHrP6LvwaQ3c03kgNhhiMgvlhxHQhmg2Xs=
+golang.org/x/crypto v0.35.0/go.mod h1:dy7dXNW32cAb/6/PRuTNsix8T+vJAqvuIy5Bli/x0YQ=
+golang.org/x/net v0.33.0 h1:74SYHlV8BIgHIFC/LrYkOGIwL19eTYXQ5wc6TBuO36I=
+golang.org/x/net v0.33.0/go.mod h1:HXLR5J+9DxmrqMwG9qjGCxZ+zKXxBru04zlTvWlWuN4=
+golang.org/x/sys v0.30.0 h1:QjkSwP/36a20jFYWkSue1YwXzLmsV5Gfq7Eiy72C1uc=
+golang.org/x/sys v0.30.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/text v0.22.0 h1:bofq7m3/HAFvbF51jz3Q9wLg3jkvSPuiZu/pD1XwgtM=
+golang.org/x/text v0.22.0/go.mod h1:YRoo4H8PVmsu+E3Ou7cqLVH8oXWIHVoX0jqUWALQhfY=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20250224174004-546df14abb99 h1:ZSlhAUqC4r8TPzqLXQ0m3upBNZeF+Y8jQ3c4CR3Ujms=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20250224174004-546df14abb99/go.mod h1:LuRYeWDFV6WOn90g357N17oMCaxpgCnbi/44qJvDn2I=
+google.golang.org/grpc v1.70.0 h1:pWFv03aZoHzlRKHWicjsZytKAiYCtNS0dHbXnIdq7jQ=
+google.golang.org/grpc v1.70.0/go.mod h1:ofIJqVKDXx/JiXrwr2IG4/zwdH9txy3IlF40RmcJSQw=
+google.golang.org/protobuf v1.36.5 h1:tPhr+woSbjfYvY6/GPufUoYizxw1cF/yFoxJ2fmpwlM=
+google.golang.org/protobuf v1.36.5/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
-gopkg.in/yaml.v3 v3.0.0/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
+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=

kgo.go

@@ -1,20 +1,25 @@
 // Package kgo provides a kafka broker using kgo
-package kgo // import "go.unistack.org/micro-broker-kgo/v3"
+package kgo
 
 import (
 	"context"
 	"errors"
 	"fmt"
-	"math/rand"
+	"math/rand/v2"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
+	"github.com/google/uuid"
 	"github.com/twmb/franz-go/pkg/kgo"
 	"github.com/twmb/franz-go/pkg/kmsg"
 	"go.unistack.org/micro/v3/broker"
+	"go.unistack.org/micro/v3/logger"
 	"go.unistack.org/micro/v3/metadata"
-	id "go.unistack.org/micro/v3/util/id"
+	"go.unistack.org/micro/v3/semconv"
+	"go.unistack.org/micro/v3/tracer"
+	mjitter "go.unistack.org/micro/v3/util/jitter"
 	mrand "go.unistack.org/micro/v3/util/rand"
 )
 
@@ -24,7 +29,6 @@ var ErrLostMessage = errors.New("message not marked for offsets commit and will
 
 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
@@ -40,7 +44,7 @@ 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)
@@ -54,12 +58,29 @@ var DefaultRetryBackoffFn = func() func(int) time.Duration {
 
 type Broker struct {
 	c         *kgo.Client
-	kopts     []kgo.Opt
-	connected bool
-	init      bool
-	sync.RWMutex
+	connected *atomic.Uint32
+
+	done chan struct{}
+
+	kopts []kgo.Opt
+	subs  []*Subscriber
+
 	opts broker.Options
-	subs []*subscriber
+	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 {
@@ -70,63 +91,119 @@ func (k *Broker) Name() string {
 	return k.opts.Name
 }
 
-func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, error) {
+func (k *Broker) Client() *kgo.Client {
+	return k.c
+}
+
+func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, *hookTracer, error) {
 	var c *kgo.Client
 	var err error
 
+	sp, _ := tracer.SpanFromContext(ctx)
+
+	clientID := "kgo"
+	group := ""
+	if k.opts.Context != nil {
+		if id, ok := k.opts.Context.Value(clientIDKey{}).(string); ok {
+			clientID = id
+		}
+		if id, ok := k.opts.Context.Value(groupKey{}).(string); ok {
+			group = id
+		}
+	}
+
+	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(htracer),
+		kgo.WithHooks(&hookEvent{log: k.opts.Logger, fatalOnError: fatalOnError, connected: k.connected}),
+	)
+
 	select {
 	case <-ctx.Done():
-		return nil, ctx.Err()
+		if ctx.Err() != nil {
+			if sp != nil {
+				sp.SetStatus(tracer.SpanStatusError, ctx.Err().Error())
+			}
+		}
+		return nil, nil, ctx.Err()
 	default:
 		c, err = kgo.NewClient(opts...)
 		if err == nil {
 			err = c.Ping(ctx) // check connectivity to cluster
 		}
 		if err != nil {
-			return nil, err
+			if sp != nil {
+				sp.SetStatus(tracer.SpanStatusError, err.Error())
+			}
+			return nil, nil, err
 		}
+		k.connected.Store(1)
+
+		if fatalOnError {
+			go func() {
+				c := 3
+				n := 0
+				tc := mjitter.NewTicker(500*time.Millisecond, 1*time.Second)
+				defer tc.Stop()
+				for range tc.C {
+					if k.connected.Load() == 0 {
+						if n > c {
+							k.opts.Logger.Fatal(context.Background(), "broker fatal error")
+						}
+						n++
+					} else {
+						n = 0
+					}
+				}
+			}()
+		}
+		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.c = c
-	k.connected = true
+	k.connected.Store(1)
 	k.Unlock()
 
 	return nil
 }
 
 func (k *Broker) Disconnect(ctx context.Context) error {
-	k.RLock()
-	if !k.connected {
-		k.RUnlock()
+	if k.connected.Load() == 0 {
 		return nil
 	}
-	k.RUnlock()
 
 	nctx := k.opts.Context
 	if ctx != nil {
 		nctx = ctx
 	}
+	var span tracer.Span
+	ctx, span = k.opts.Tracer.Start(ctx, "Disconnect")
+	defer span.Finish()
 
 	k.Lock()
 	defer k.Unlock()
@@ -148,7 +225,8 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 		}
 	}
 
-	k.connected = false
+	k.connected.Store(0)
+	close(k.done)
 	return nil
 }
 
@@ -156,6 +234,10 @@ func (k *Broker) Init(opts ...broker.Option) error {
 	k.Lock()
 	defer k.Unlock()
 
+	if len(opts) == 0 && k.init {
+		return nil
+	}
+
 	for _, o := range opts {
 		o(&k.opts)
 	}
@@ -179,6 +261,8 @@ func (k *Broker) Init(opts ...broker.Option) error {
 		}
 	}
 
+	k.init = true
+
 	return nil
 }
 
@@ -196,43 +280,32 @@ func (k *Broker) Publish(ctx context.Context, topic string, msg *broker.Message,
 }
 
 func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...broker.PublishOption) error {
-	k.RLock()
-	if !k.connected {
-		k.RUnlock()
-		k.Lock()
-		c, err := k.connect(ctx, k.kopts...)
-		if err != nil {
-			k.Unlock()
-			return err
-		}
-		k.c = c
-		k.connected = true
-		k.Unlock()
-	}
-	k.RUnlock()
-
 	options := broker.NewPublishOptions(opts...)
 	records := make([]*kgo.Record, 0, len(msgs))
 	var errs []string
 	var err error
 	var key []byte
+	var promise func(*kgo.Record, error)
 
 	if options.Context != nil {
 		if k, ok := options.Context.Value(publishKey{}).([]byte); ok && k != nil {
 			key = k
 		}
+		if p, ok := options.Context.Value(publishPromiseKey{}).(func(*kgo.Record, error)); ok && p != nil {
+			promise = p
+		}
 	}
 
 	for _, msg := range msgs {
 		rec := &kgo.Record{Context: ctx, Key: key}
+
 		rec.Topic, _ = msg.Header.Get(metadata.HeaderTopic)
-		if options.BodyOnly {
+		msg.Header.Del(metadata.HeaderTopic)
+
+		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
-		} else if 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)})
-			}
+			setHeaders(rec, msg.Header)
 		} else {
 			rec.Value, err = k.opts.Codec.Marshal(msg)
 			if err != nil {
@@ -242,10 +315,38 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 		records = append(records, rec)
 	}
 
+	if promise != nil {
+		ts := time.Now()
+		for _, rec := range records {
+			k.c.Produce(ctx, rec, func(r *kgo.Record, err error) {
+				te := time.Since(ts)
+				k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", rec.Topic, "topic", rec.Topic).Dec()
+				k.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", rec.Topic, "topic", rec.Topic).Update(te.Seconds())
+				k.opts.Meter.Histogram(semconv.PublishMessageDurationSeconds, "endpoint", rec.Topic, "topic", rec.Topic).Update(te.Seconds())
+				if err != nil {
+					k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "failure").Inc()
+				} else {
+					k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", rec.Topic, "topic", rec.Topic, "status", "success").Inc()
+				}
+				promise(r, err)
+			})
+		}
+		return nil
+	}
+	ts := time.Now()
+
 	results := k.c.ProduceSync(ctx, records...)
+
+	te := time.Since(ts)
 	for _, result := range results {
+		k.opts.Meter.Summary(semconv.PublishMessageLatencyMicroseconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+		k.opts.Meter.Histogram(semconv.PublishMessageDurationSeconds, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Update(te.Seconds())
+		k.opts.Meter.Counter(semconv.PublishMessageInflight, "endpoint", result.Record.Topic, "topic", result.Record.Topic).Dec()
 		if result.Err != nil {
+			k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "failure").Inc()
 			errs = append(errs, result.Err.Error())
+		} else {
+			k.opts.Meter.Counter(semconv.PublishMessageTotal, "endpoint", result.Record.Topic, "topic", result.Record.Topic, "status", "success").Inc()
 		}
 	}
 
@@ -256,7 +357,23 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 	return nil
 }
 
-func (k *Broker) BatchSubscribe(ctx context.Context, topic string, handler broker.BatchHandler, opts ...broker.SubscribeOption) (broker.Subscriber, error) {
+func (k *Broker) TopicExists(ctx context.Context, topic string) error {
+	mdreq := kmsg.NewMetadataRequest()
+	mdreq.Topics = []kmsg.MetadataRequestTopic{
+		{Topic: &topic},
+	}
+
+	mdrsp, err := mdreq.RequestWith(ctx, k.c)
+	if err != nil {
+		return err
+	} else if mdrsp.Topics[0].ErrorCode != 0 {
+		return fmt.Errorf("topic %s not exists or permission error", topic)
+	}
+
+	return nil
+}
+
+func (k *Broker) BatchSubscribe(_ context.Context, _ string, _ broker.BatchHandler, _ ...broker.SubscribeOption) (broker.Subscriber, error) {
 	return nil, nil
 }
 
@@ -264,11 +381,11 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	options := broker.NewSubscribeOptions(opts...)
 
 	if options.Group == "" {
-		uid, err := id.New()
+		uid, err := uuid.NewRandom()
 		if err != nil {
 			return nil, err
 		}
-		options.Group = uid
+		options.Group = uid.String()
 	}
 
 	commitInterval := DefaultCommitInterval
@@ -278,13 +395,22 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	sub := &subscriber{
-		topic:     topic,
-		opts:      options,
-		handler:   handler,
-		kopts:     k.opts,
-		consumers: make(map[tp]*consumer),
-		done:      make(chan struct{}),
+	var fatalOnError bool
+	if k.opts.Context != nil {
+		if v, ok := k.opts.Context.Value(fatalOnErrorKey{}).(bool); ok && v {
+			fatalOnError = v
+		}
+	}
+
+	sub := &Subscriber{
+		topic:        topic,
+		opts:         options,
+		handler:      handler,
+		kopts:        k.opts,
+		consumers:    make(map[tp]*consumer),
+		done:         make(chan struct{}),
+		fatalOnError: fatalOnError,
+		connected:    k.connected,
 	}
 
 	kopts := append(k.kopts,
@@ -295,7 +421,9 @@ 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(),
 	)
 
@@ -305,7 +433,7 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
-	c, err := k.connect(ctx, kopts...)
+	c, htracer, err := k.connect(ctx, kopts...)
 	if err != nil {
 		return nil, err
 	}
@@ -323,6 +451,8 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 	}
 
 	sub.c = c
+	sub.htracer = htracer
+
 	go sub.poll(ctx)
 
 	k.Lock()
@@ -336,7 +466,6 @@ func (k *Broker) String() string {
 }
 
 func NewBroker(opts ...broker.Option) *Broker {
-	rand.Seed(time.Now().Unix())
 	options := broker.NewOptions(opts...)
 
 	kaddrs := options.Addrs
@@ -349,14 +478,13 @@ func NewBroker(opts ...broker.Option) *Broker {
 		kgo.DialTimeout(3 * time.Second),
 		kgo.DisableIdempotentWrite(),
 		kgo.ProducerBatchCompression(kgo.NoCompression()),
-		kgo.WithLogger(&mlogger{l: options.Logger, ctx: options.Context}),
-		// kgo.WithLogger(kgo.BasicLogger(os.Stderr, kgo.LogLevelDebug, func() string { return time.Now().Format(time.StampMilli) })),
-		kgo.WithHooks(&metrics{meter: options.Meter}),
+		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 {
@@ -366,7 +494,9 @@ func NewBroker(opts ...broker.Option) *Broker {
 	}
 
 	return &Broker{
-		opts:  options,
-		kopts: kopts,
+		connected: &atomic.Uint32{},
+		opts:      options,
+		kopts:     kopts,
+		done:      make(chan struct{}),
 	}
 }

kgo_test.go

@@ -2,7 +2,6 @@ package kgo_test
 
 import (
 	"context"
-	"fmt"
 	"os"
 	"strings"
 	"sync/atomic"
@@ -13,14 +12,15 @@ import (
 	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/logger/slog"
 	"go.unistack.org/micro/v3/metadata"
 )
 
 var (
-	msgcnt   = int64(12000000)
+	msgcnt   = int64(1200)
 	group    = "38"
-	prefill  = false
-	loglevel = logger.InfoLevel
+	prefill  = true
+	loglevel = logger.DebugLevel
 )
 
 var bm = &broker.Message{
@@ -28,6 +28,94 @@ var bm = &broker.Message{
 	Body:   []byte(`"body"`),
 }
 
+func TestFail(t *testing.T) {
+	if tr := os.Getenv("INTEGRATION_TESTS"); len(tr) > 0 {
+		t.Skip()
+	}
+
+	logger.DefaultLogger = slog.NewLogger()
+	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel)); err != nil {
+		t.Fatal(err)
+	}
+	ctx := context.Background()
+
+	var addrs []string
+	if addr := os.Getenv("BROKER_ADDRS"); len(addr) == 0 {
+		addrs = []string{"127.0.0.1:9092"}
+	} else {
+		addrs = strings.Split(addr, ",")
+	}
+
+	b := kgo.NewBroker(
+		broker.Addrs(addrs...),
+		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++ {
+		msgs = append(msgs, bm)
+	}
+
+	for _, msg := range msgs {
+		t.Logf("broker publish")
+		if err := b.Publish(ctx, "test", msg); err != nil {
+			t.Fatal(err)
+		}
+	}
+	//	t.Skip()
+
+	idx := int64(0)
+	fn := func(msg broker.Event) error {
+		atomic.AddInt64(&idx, 1)
+		time.Sleep(500 * time.Millisecond)
+		t.Logf("ack")
+		return msg.Ack()
+	}
+
+	sub, err := b.Subscribe(ctx, "test", 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
+		}
+		time.Sleep(1 * time.Second)
+	}
+}
+
 func TestConnect(t *testing.T) {
 	var addrs []string
 	ctx := context.TODO()
@@ -50,7 +138,7 @@ func TestPubSub(t *testing.T) {
 		t.Skip()
 	}
 
-	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel), logger.WithCallerSkipCount(3)); err != nil {
+	if err := logger.DefaultLogger.Init(logger.WithLevel(loglevel)); err != nil {
 		t.Fatal(err)
 	}
 	ctx := context.Background()
@@ -67,6 +155,7 @@ func TestPubSub(t *testing.T) {
 		kgo.CommitInterval(5*time.Second),
 		kgo.Options(kg.ClientID("test"), kg.FetchMaxBytes(10*1024*1024)),
 	)
+
 	if err := b.Init(); err != nil {
 		t.Fatal(err)
 	}
@@ -124,7 +213,7 @@ 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\n", prc)
 				}
 			case <-ticker.C:
 				close(done)

logger.go

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

meter.go

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

options.go

@@ -9,8 +9,17 @@ import (
 	"go.unistack.org/micro/v3/client"
 )
 
-// DefaultCommitInterval specifies how fast send commit offsets to kafka
-var DefaultCommitInterval = 5 * time.Second
+var (
+
+	// DefaultCommitInterval specifies how fast send commit offsets to kafka
+	DefaultCommitInterval = 5 * time.Second
+
+	// DefaultStatsInterval specifies how fast check consumer lag
+	DefaultStatsInterval = 5 * time.Second
+
+	// DefaultSubscribeMaxInflight specifies how much messages keep inflight
+	DefaultSubscribeMaxInflight = 100
+)
 
 type subscribeContextKey struct{}
 
@@ -63,6 +72,24 @@ 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 {
+	return broker.SetOption(clientIDKey{}, id)
+}
+
+type groupKey struct{}
+
+func Group(id string) broker.Option {
+	return broker.SetOption(groupKey{}, id)
+}
+
 type commitIntervalKey struct{}
 
 // CommitInterval specifies interval to send commits
@@ -70,11 +97,26 @@ func CommitInterval(td time.Duration) broker.Option {
 	return broker.SetOption(commitIntervalKey{}, td)
 }
 
-var DefaultSubscribeMaxInflight = 10
-
 type subscribeMaxInflightKey struct{}
 
 // SubscribeMaxInFlight max queued messages
 func SubscribeMaxInFlight(n int) broker.SubscribeOption {
 	return broker.SetSubscribeOption(subscribeMaxInflightKey{}, n)
 }
+
+// SubscribeMaxInFlight max queued messages
+func SubscribeFatalOnError(b bool) broker.SubscribeOption {
+	return broker.SetSubscribeOption(fatalOnErrorKey{}, b)
+}
+
+type publishPromiseKey struct{}
+
+// PublishPromise set the kafka promise func for Produce
+func PublishPromise(fn func(*kgo.Record, error)) broker.PublishOption {
+	return broker.SetPublishOption(publishPromiseKey{}, fn)
+}
+
+// ClientPublishKey set the kafka message key (client option)
+func ClientPublishPromise(fn func(*kgo.Record, error)) client.PublishOption {
+	return client.SetPublishOption(publishPromiseKey{}, fn)
+}

subscriber.go

@@ -2,12 +2,20 @@ package kgo
 
 import (
 	"context"
+	"fmt"
+	"strconv"
 	"sync"
+	"sync/atomic"
+	"time"
 
+	"github.com/twmb/franz-go/pkg/kadm"
 	"github.com/twmb/franz-go/pkg/kgo"
+	"github.com/twmb/franz-go/pkg/kmsg"
 	"go.unistack.org/micro/v3/broker"
 	"go.unistack.org/micro/v3/logger"
 	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/semconv"
+	"go.unistack.org/micro/v3/tracer"
 )
 
 type tp struct {
@@ -16,65 +24,119 @@ type tp struct {
 }
 
 type consumer struct {
+	topic string
+
 	c         *kgo.Client
-	topic     string
+	htracer   *hookTracer
+	connected *atomic.Uint32
+
+	quit chan struct{}
+	done chan struct{}
+	recs chan kgo.FetchTopicPartition
+
+	handler broker.Handler
+
+	kopts broker.Options
+	opts  broker.SubscribeOptions
+
 	partition int32
-	opts      broker.SubscribeOptions
-	kopts     broker.Options
-	handler   broker.Handler
-	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{}
+type Subscriber struct {
+	topic string
+
 	consumers map[tp]*consumer
+
+	c         *kgo.Client
+	htracer   *hookTracer
+	connected *atomic.Uint32
+
+	handler broker.Handler
+
+	done chan struct{}
+
+	kopts broker.Options
+	opts  broker.SubscribeOptions
+
+	closed       bool
+	fatalOnError bool
+
 	sync.RWMutex
 }
 
-func (s *subscriber) Options() broker.SubscribeOptions {
+func (s *Subscriber) Client() *kgo.Client {
+	return s.c
+}
+
+func (s *Subscriber) Options() broker.SubscribeOptions {
 	return s.opts
 }
 
-func (s *subscriber) Topic() string {
+func (s *Subscriber) Topic() string {
 	return s.topic
 }
 
-func (s *subscriber) Unsubscribe(ctx context.Context) error {
+func (s *Subscriber) Unsubscribe(ctx context.Context) error {
 	if s.closed {
 		return nil
 	}
-	select {
-	// case <-ctx.Done():
-	//	return ctx.Err()
-	default:
-		s.c.PauseFetchTopics(s.topic)
-		s.c.CloseAllowingRebalance()
-		kc := make(map[string][]int32)
-		for ctp := range s.consumers {
-			kc[ctp.t] = append(kc[ctp.t], ctp.p)
-		}
-		s.killConsumers(ctx, kc)
-		close(s.done)
-		s.closed = true
+
+	s.c.PauseFetchTopics(s.topic)
+	s.c.CloseAllowingRebalance()
+	kc := make(map[string][]int32)
+	for ctp := range s.consumers {
+		kc[ctp.t] = append(kc[ctp.t], ctp.p)
 	}
+	s.killConsumers(ctx, kc)
+	close(s.done)
+	s.closed = true
+	s.c.ResumeFetchTopics(s.topic)
+
 	return nil
 }
 
-func (s *subscriber) poll(ctx context.Context) {
+func (s *Subscriber) poll(ctx context.Context) {
 	maxInflight := DefaultSubscribeMaxInflight
 	if s.opts.Context != nil {
 		if n, ok := s.opts.Context.Value(subscribeMaxInflightKey{}).(int); n > 0 && ok {
 			maxInflight = n
 		}
 	}
+
+	go func() {
+		ac := kadm.NewClient(s.c)
+		ticker := time.NewTicker(DefaultStatsInterval)
+
+		for {
+			select {
+			case <-ctx.Done():
+				ticker.Stop()
+				return
+			case <-ticker.C:
+				dgls, err := ac.Lag(ctx, s.opts.Group)
+				if err != nil || !dgls.Ok() {
+					continue
+				}
+
+				dgl, ok := dgls[s.opts.Group]
+				if !ok {
+					continue
+				}
+				lmap, ok := dgl.Lag[s.topic]
+				if !ok {
+					continue
+				}
+
+				s.Lock()
+				for p, l := range lmap {
+					s.kopts.Meter.Counter(semconv.BrokerGroupLag, "topic", s.topic, "group", s.opts.Group, "partition", strconv.Itoa(int(p))).Set(uint64(l.Lag))
+				}
+				s.Unlock()
+
+			}
+		}
+	}()
+
 	for {
 		select {
 		case <-ctx.Done():
@@ -89,64 +151,87 @@ func (s *subscriber) poll(ctx context.Context) {
 				return
 			}
 			fetches.EachError(func(t string, p int32, err error) {
-				s.kopts.Logger.Fatalf(ctx, "[kgo] fetch topic %s partition %d err: %v", t, p, err)
+				s.kopts.Logger.Fatal(ctx, fmt.Sprintf("[kgo] fetch topic %s partition %d error", t, p), err)
 			})
 
 			fetches.EachPartition(func(p kgo.FetchTopicPartition) {
-				tp := tp{p.Topic, p.Partition}
-				s.consumers[tp].recs <- p
+				tps := tp{p.Topic, p.Partition}
+				s.consumers[tps].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}
+			pc, ok := s.consumers[tps]
+			if !ok {
+				continue
+			}
+			delete(s.consumers, tps)
 			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) revoked(ctx context.Context, c *kgo.Client, revoked map[string][]int32) {
-	s.kopts.Logger.Debugf(ctx, "[kgo] revoked %#+v", revoked)
-	s.killConsumers(ctx, revoked)
-	if err := c.CommitMarkedOffsets(ctx); err != nil {
-		s.kopts.Logger.Errorf(ctx, "[kgo] revoked CommitMarkedOffsets err: %v", err)
+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) assigned(_ context.Context, c *kgo.Client, assigned map[string][]int32) {
+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.Error(ctx, "[kgo] revoked CommitMarkedOffsets error", err)
+		// s.connected.Store(0)
+	}
+}
+
+func (s *Subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[string][]int32) {
 	for topic, partitions := range assigned {
 		for _, partition := range partitions {
 			pc := &consumer{
 				c:         c,
 				topic:     topic,
 				partition: partition,
-
-				quit:    make(chan struct{}),
-				done:    make(chan struct{}),
-				recs:    make(chan kgo.FetchTopicPartition, 4),
-				handler: s.handler,
-				kopts:   s.kopts,
-				opts:    s.opts,
+				htracer:   s.htracer,
+				quit:      make(chan struct{}),
+				done:      make(chan struct{}),
+				recs:      make(chan kgo.FetchTopicPartition, 100),
+				handler:   s.handler,
+				kopts:     s.kopts,
+				opts:      s.opts,
+				connected: s.connected,
 			}
+			s.Lock()
 			s.consumers[tp{topic, partition}] = pc
+			s.Unlock()
 			go pc.consume()
 		}
 	}
@@ -154,8 +239,10 @@ func (s *subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[str
 
 func (pc *consumer) consume() {
 	defer close(pc.done)
-	pc.kopts.Logger.Debugf(pc.kopts.Context, "starting, topic %s partition %d", pc.topic, pc.partition)
-	defer pc.kopts.Logger.Debugf(pc.kopts.Context, "killing, topic %s partition %d", pc.topic, pc.partition)
+	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))
+	}
 
 	eh := pc.kopts.ErrorHandler
 	if pc.opts.ErrorHandler != nil {
@@ -168,66 +255,124 @@ func (pc *consumer) consume() {
 			return
 		case p := <-pc.recs:
 			for _, record := range p.Records {
+				ctx, sp := pc.htracer.WithProcessSpan(record)
+				ts := time.Now()
+				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Inc()
 				p := eventPool.Get().(*event)
 				p.msg.Header = nil
 				p.msg.Body = nil
 				p.topic = record.Topic
 				p.err = nil
 				p.ack = false
+				p.msg.Header = metadata.New(len(record.Headers))
+				p.ctx = ctx
+				for _, hdr := range record.Headers {
+					p.msg.Header.Set(hdr.Key, string(hdr.Value))
+				}
 				if pc.kopts.Codec.String() == "noop" {
-					p.msg.Header = metadata.New(len(record.Headers))
-					for _, hdr := range record.Headers {
-						p.msg.Header.Set(hdr.Key, string(hdr.Value))
-					}
 					p.msg.Body = record.Value
 				} else if pc.opts.BodyOnly {
 					p.msg.Body = record.Value
 				} else {
-					if err := pc.kopts.Codec.Unmarshal(record.Value, p.msg); err != nil {
+					if sp != nil {
+						sp.AddEvent("codec unmarshal start")
+					}
+					err := pc.kopts.Codec.Unmarshal(record.Value, p.msg)
+					if sp != nil {
+						sp.AddEvent("codec unmarshal stop")
+					}
+					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()
 						p.err = err
 						p.msg.Body = record.Value
 						if eh != nil {
 							_ = eh(p)
+							pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 							if p.ack {
 								pc.c.MarkCommitRecords(record)
 							} else {
 								eventPool.Put(p)
-								pc.kopts.Logger.Infof(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+								//								pc.connected.Store(0)
+								pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
 								return
 							}
 							eventPool.Put(p)
+							te := time.Since(ts)
+							pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+							pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 							continue
 						} else {
-							if pc.kopts.Logger.V(logger.ErrorLevel) {
-								pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: failed to unmarshal: %v", err)
-							}
+							pc.kopts.Logger.Error(pc.kopts.Context, "[kgo]: unmarshal error", err)
 						}
+						te := time.Since(ts)
+						pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
+						pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+						pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 						eventPool.Put(p)
-						pc.kopts.Logger.Infof(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
+						//						pc.connected.Store(0)
+						pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
+						if sp != nil {
+							sp.Finish()
+						}
 						return
 					}
 				}
+				if sp != nil {
+					sp.AddEvent("handler start")
+				}
 				err := pc.handler(p)
+				if sp != nil {
+					sp.AddEvent("handler stop")
+				}
+				if err == nil {
+					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "success").Inc()
+				} else {
+					if sp != nil {
+						sp.SetStatus(tracer.SpanStatusError, err.Error())
+					}
+					pc.kopts.Meter.Counter(semconv.SubscribeMessageTotal, "endpoint", record.Topic, "topic", record.Topic, "status", "failure").Inc()
+				}
+				pc.kopts.Meter.Counter(semconv.SubscribeMessageInflight, "endpoint", record.Topic, "topic", record.Topic).Dec()
 				if err == nil && pc.opts.AutoAck {
 					p.ack = true
 				} else if err != nil {
 					p.err = err
 					if eh != nil {
+						if sp != nil {
+							sp.AddEvent("error handler start")
+						}
 						_ = eh(p)
+						if sp != nil {
+							sp.AddEvent("error handler stop")
+						}
 					} else {
 						if pc.kopts.Logger.V(logger.ErrorLevel) {
-							pc.kopts.Logger.Errorf(pc.kopts.Context, "[kgo]: subscriber error: %v", err)
+							pc.kopts.Logger.Error(pc.kopts.Context, "[kgo]: subscriber error", err)
 						}
 					}
 				}
+				te := time.Since(ts)
+				pc.kopts.Meter.Summary(semconv.SubscribeMessageLatencyMicroseconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
+				pc.kopts.Meter.Histogram(semconv.SubscribeMessageDurationSeconds, "endpoint", record.Topic, "topic", record.Topic).Update(te.Seconds())
 				if p.ack {
 					eventPool.Put(p)
 					pc.c.MarkCommitRecords(record)
 				} else {
 					eventPool.Put(p)
-					pc.kopts.Logger.Fatalf(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+					//					pc.connected.Store(0)
+					pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
+					if sp != nil {
+						sp.SetStatus(tracer.SpanStatusError, "ErrLostMessage")
+						sp.Finish()
+					}
 					return
 				}
+				if sp != nil {
+					sp.Finish()
+				}
 			}
 		}
 	}

tracer.go

@@ -0,0 +1,223 @@
+package kgo
+
+import (
+	"context"
+	"unicode/utf8"
+
+	"github.com/twmb/franz-go/pkg/kgo"
+	semconv "go.opentelemetry.io/otel/semconv/v1.18.0"
+	"go.unistack.org/micro/v3/metadata"
+	"go.unistack.org/micro/v3/tracer"
+)
+
+type hookTracer struct {
+	tracer   tracer.Tracer
+	clientID string
+	group    string
+}
+
+var messagingSystem = semconv.MessagingSystemKey.String("kafka")
+
+var (
+	_ kgo.HookProduceRecordBuffered   = (*hookTracer)(nil)
+	_ kgo.HookProduceRecordUnbuffered = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordBuffered     = (*hookTracer)(nil)
+	_ kgo.HookFetchRecordUnbuffered   = (*hookTracer)(nil)
+)
+
+// OnProduceRecordBuffered starts a new span for the "publish" operation on a
+// buffered record.
+//
+// It sets span options and injects the span context into record and updates
+// the record's context, so it can be ended in the OnProduceRecordUnbuffered
+// hook.
+func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
+	// Set up span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingDestinationKindTopic,
+		semconv.MessagingDestinationName(r.Topic),
+		semconv.MessagingOperationPublish,
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	opts := []tracer.SpanOption{
+		tracer.WithSpanLabels(attrs...),
+		tracer.WithSpanKind(tracer.SpanKindProducer),
+	}
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	omd, ok := metadata.FromOutgoingContext(r.Context)
+	if !ok {
+		omd = metadata.New(len(r.Headers))
+	}
+
+	md := metadata.Copy(omd)
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewOutgoingContext(r.Context, md), "sdk.broker", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
+	}
+
+	setHeaders(r, omd, metadata.HeaderContentType)
+}
+
+// OnProduceRecordUnbuffered continues and ends the "publish" span for an
+// unbuffered record.
+//
+// It sets attributes with values unset when producing and records any error
+// that occurred during the publish operation.
+func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
+	if !m.tracer.Enabled() {
+		return
+	}
+	if span, ok := tracer.SpanFromContext(r.Context); ok {
+		span.AddLabels(
+			semconv.MessagingKafkaDestinationPartition(int(r.Partition)),
+		)
+		if err != nil {
+			span.SetStatus(tracer.SpanStatusError, err.Error())
+		}
+		span.Finish()
+	}
+}
+
+// OnFetchRecordBuffered starts a new span for the "receive" operation on a
+// buffered record.
+//
+// It sets the span options and extracts the span context from the record,
+// updates the record's context to ensure it can be ended in the
+// OnFetchRecordUnbuffered hook and can be used in downstream consumer
+// processing.
+func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
+	// Set up the span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingSourceKindTopic,
+		semconv.MessagingSourceName(r.Topic),
+		semconv.MessagingOperationReceive,
+		semconv.MessagingKafkaSourcePartition(int(r.Partition)),
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	if m.group != "" {
+		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
+	}
+	opts := []tracer.SpanOption{
+		tracer.WithSpanLabels(attrs...),
+		tracer.WithSpanKind(tracer.SpanKindConsumer),
+	}
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+	omd, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		omd = metadata.New(len(r.Headers))
+	}
+
+	md := metadata.Copy(omd)
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
+	if !ok {
+		r.Context, _ = m.tracer.Start(metadata.NewIncomingContext(r.Context, md), "sdk.broker", opts...)
+	} else {
+		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
+	}
+
+	setHeaders(r, omd, metadata.HeaderContentType)
+}
+
+// OnFetchRecordUnbuffered continues and ends the "receive" span for an
+// unbuffered record.
+func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
+	if !m.tracer.Enabled() {
+		return
+	}
+	span, _ := tracer.SpanFromContext(r.Context)
+	span.Finish()
+}
+
+// WithProcessSpan starts a new span for the "process" operation on a consumer
+// record.
+//
+// It sets up the span options. The user's application code is responsible for
+// ending the span.
+//
+// This should only ever be called within a polling loop of a consumed record and
+// not a record which has been created for producing, so call this at the start of each
+// iteration of your processing for the record.
+func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Span) {
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	if !m.tracer.Enabled() {
+		return r.Context, nil
+	}
+	// Set up the span options.
+	attrs := []interface{}{
+		messagingSystem,
+		semconv.MessagingSourceKindTopic,
+		semconv.MessagingSourceName(r.Topic),
+		semconv.MessagingOperationProcess,
+		semconv.MessagingKafkaSourcePartition(int(r.Partition)),
+		semconv.MessagingKafkaMessageOffset(int(r.Offset)),
+	}
+	attrs = maybeKeyAttr(attrs, r)
+	if m.clientID != "" {
+		attrs = append(attrs, semconv.MessagingKafkaClientIDKey.String(m.clientID))
+	}
+	if m.group != "" {
+		attrs = append(attrs, semconv.MessagingKafkaConsumerGroupKey.String(m.group))
+	}
+	opts := []tracer.SpanOption{
+		tracer.WithSpanLabels(attrs...),
+		tracer.WithSpanKind(tracer.SpanKindConsumer),
+	}
+
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+	md, ok := metadata.FromIncomingContext(r.Context)
+	if !ok {
+		md = metadata.New(len(r.Headers))
+	}
+	for _, h := range r.Headers {
+		md.Set(h.Key, string(h.Value))
+	}
+
+	// Start a new span using the provided context and options.
+	return m.tracer.Start(r.Context, "sdk.broker", opts...)
+}
+
+func maybeKeyAttr(attrs []interface{}, r *kgo.Record) []interface{} {
+	if r.Key == nil {
+		return attrs
+	}
+	var keykey string
+	if !utf8.Valid(r.Key) {
+		return attrs
+	}
+	keykey = string(r.Key)
+	return append(attrs, semconv.MessagingKafkaMessageKeyKey.String(keykey))
+}