update deps

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

.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

@@ -3,10 +3,10 @@ name: lint
 on:
   pull_request:
     types: [opened, reopened, synchronize]
-    branches:
+    branches: [ master, v3, v4 ]
-    - master
+    paths-ignore:
-    - v3
+      - '.github/**'
-    - v4
+      - '.gitea/**'
 
 jobs:
   lint:
@@ -20,10 +20,10 @@ jobs:
       uses: actions/setup-go@v5
       with:
         cache-dependency-path: "**/*.sum"
-        go-version: 'stable' 
+        go-version: 'stable'
     - name: setup deps
       run: go get -v ./...
     - name: run lint
-      uses: https://github.com/golangci/golangci-lint-action@v6
+      uses: golangci/golangci-lint-action@v6
       with:
         version: 'latest'

.github/workflows/job_test.yml

@@ -3,15 +3,12 @@ name: test
 on:
   pull_request:
     types: [opened, reopened, synchronize]
-    branches:
+    branches: [ master, v3, v4 ]
-    - master
-    - v3
-    - v4
   push:
-    branches:
+    branches: [ master, v3, v4 ]
-    - master
+    paths-ignore:
-    - v3
+      - '.github/**'
-    - v4
+      - '.gitea/**'
 
 jobs:
   test:

.github/workflows/job_tests.yml

@@ -3,15 +3,12 @@ name: test
 on:
   pull_request:
     types: [opened, reopened, synchronize]
-    branches:
+    branches: [ master, v3, v4 ]
-    - master
-    - v3
-    - v4
   push:
-    branches:
+    branches: [ master, v3, v4 ]
-    - master
+    paths-ignore:
-    - v3
+      - '.github/**'
-    - v4
+      - '.gitea/**'
 
 jobs:
   test:
@@ -35,19 +32,19 @@ jobs:
         go-version: 'stable'
     - name: setup go work
       env:
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: |
         go work init
         go work use .
-        go work use micro-tests        
+        go work use micro-tests
     - name: setup deps
       env:
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: go get -v ./...
     - name: run tests
       env:
         INTEGRATION_TESTS: yes
-        GOWORK: /workspace/${{ github.repository_owner }}/go.work
+        GOWORK: ${{ github.workspace }}/go.work
       run: |
         cd micro-tests
-        go test -mod readonly -v ./... || true        
+        go test -mod readonly -v ./... || true

.golangci.yml

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

carrier.go

@@ -1,6 +1,10 @@
 package kgo
 
 import (
+	"net/http"
+	"slices"
+	"strings"
+
 	"github.com/twmb/franz-go/pkg/kgo"
 	"go.unistack.org/micro/v3/metadata"
 )
@@ -53,24 +57,37 @@ func (c RecordCarrier) Keys() []string {
 	return out
 }
 
-func setHeaders(r *kgo.Record, md metadata.Metadata) {
+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 r.Headers[i].Key == k {
+			if strings.EqualFold(r.Headers[i].Key, k) {
 				// Key exist, update the value.
 				r.Headers[i].Value = []byte(v)
 				continue loop
-			} else if _, ok := seen[k]; ok {
-				continue loop
 			}
-			// Key does not exist, append new header.
-			r.Headers = append(r.Headers, kgo.RecordHeader{
-				Key:   k,
-				Value: []byte(v),
-			})
-			seen[k] = struct{}{}
 		}
+
+		// Key does not exist, append new header.
+		r.Headers = append(r.Headers, kgo.RecordHeader{
+			Key:   k,
+			Value: []byte(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)
+}

go.mod

@@ -1,27 +1,32 @@
 module go.unistack.org/micro-broker-kgo/v3
 
-go 1.22.0
+go 1.24.0
+
+toolchain go1.24.3
 
 require (
 	github.com/google/uuid v1.6.0
-	github.com/twmb/franz-go v1.18.0
+	github.com/stretchr/testify v1.11.1
-	github.com/twmb/franz-go/pkg/kadm v1.14.0
+	github.com/twmb/franz-go v1.20.2
-	github.com/twmb/franz-go/pkg/kmsg v1.9.0
+	github.com/twmb/franz-go/pkg/kadm v1.17.1
-	go.opentelemetry.io/otel v1.33.0
+	github.com/twmb/franz-go/pkg/kmsg v1.12.0
-	go.unistack.org/micro/v3 v3.11.37
+	go.opentelemetry.io/otel v1.38.0
+	go.unistack.org/micro/v3 v3.11.48
 )
 
 require (
 	github.com/ash3in/uuidv8 v1.2.0 // indirect
-	github.com/klauspost/compress v1.17.11 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
+	github.com/klauspost/compress v1.18.1 // 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
+	github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 // indirect
 	go.unistack.org/micro-proto/v3 v3.4.1 // indirect
-	golang.org/x/crypto v0.31.0 // indirect
+	golang.org/x/crypto v0.43.0 // indirect
-	golang.org/x/sys v0.28.0 // indirect
+	golang.org/x/sys v0.37.0 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20241216192217-9240e9c98484 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20251022142026-3a174f9686a8 // indirect
-	google.golang.org/grpc v1.69.2 // indirect
+	google.golang.org/grpc v1.76.0 // indirect
-	google.golang.org/protobuf v1.36.1 // indirect
+	google.golang.org/protobuf v1.36.10 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -7,12 +7,14 @@ github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1
 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 v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
-github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
+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 v1.17.11 h1:In6xLpyWOi1+C7tXUUWv2ot1QvBjxevKAaI6IXrJmUc=
+github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
-github.com/klauspost/compress v1.17.11/go.mod h1:pMDklpSncoRMuLFrf1W9Ss9KT+0rH90U12bZKk7uwG0=
+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=
@@ -27,34 +29,44 @@ github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2/go.mod h1:iKH
 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 v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
-github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
+github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
-github.com/twmb/franz-go v1.18.0 h1:25FjMZfdozBywVX+5xrWC2W+W76i0xykKjTdEeD2ejw=
+github.com/twmb/franz-go v1.19.5 h1:W7+o8D0RsQsedqib71OVlLeZ0zI6CbFra7yTYhZTs5Y=
-github.com/twmb/franz-go v1.18.0/go.mod h1:zXCGy74M0p5FbXsLeASdyvfLFsBvTubVqctIaa5wQ+I=
+github.com/twmb/franz-go v1.19.5/go.mod h1:4kFJ5tmbbl7asgwAGVuyG1ZMx0NNpYk7EqflvWfPCpM=
-github.com/twmb/franz-go/pkg/kadm v1.14.0 h1:nAn1co1lXzJQocpzyIyOFOjUBf4WHWs5/fTprXy2IZs=
+github.com/twmb/franz-go v1.20.2 h1:CiwhyKZHW6vqSHJkh+RTxFAJkio0jBjM/JQhx/HZ72A=
-github.com/twmb/franz-go/pkg/kadm v1.14.0/go.mod h1:XjOPz6ZaXXjrW2jVCfLuucP8H1w2TvD6y3PT2M+aAM4=
+github.com/twmb/franz-go v1.20.2/go.mod h1:YCnepDd4gl6vdzG03I5Wa57RnCTIC6DVEyMpDX/J8UA=
-github.com/twmb/franz-go/pkg/kmsg v1.9.0 h1:JojYUph2TKAau6SBtErXpXGC7E3gg4vGZMv9xFU/B6M=
+github.com/twmb/franz-go/pkg/kadm v1.16.1 h1:IEkrhTljgLHJ0/hT/InhXGjPdmWfFvxp7o/MR7vJ8cw=
-github.com/twmb/franz-go/pkg/kmsg v1.9.0/go.mod h1:CMbfazviCyY6HM0SXuG5t9vOwYDHRCSrJJyBAe5paqg=
+github.com/twmb/franz-go/pkg/kadm v1.16.1/go.mod h1:Ue/ye1cc9ipsQFg7udFbbGiFNzQMqiH73fGC2y0rwyc=
-go.opentelemetry.io/otel v1.33.0 h1:/FerN9bax5LoK51X/sI0SVYrjSE0/yUL7DpxW4K3FWw=
+github.com/twmb/franz-go/pkg/kadm v1.17.1 h1:Bt02Y/RLgnFO2NP2HVP1kd2TFtGRiJZx+fSArjZDtpw=
-go.opentelemetry.io/otel v1.33.0/go.mod h1:SUUkR6csvUQl+yjReHu5uM3EtVV7MBm5FHKRlNx4I8I=
+github.com/twmb/franz-go/pkg/kadm v1.17.1/go.mod h1:s4duQmrDbloVW9QTMXhs6mViTepze7JLG43xwPcAeTg=
+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/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 v3.11.37 h1:ZcpnXAYEMcAwmnVb5b7o8/PylGnILxXMHaUlRrPmRI0=
+go.unistack.org/micro/v3 v3.11.45 h1:fjTLZYWgsVf9FIMZBxOg8ios2/tmyimnjZrsrxEUeXU=
-go.unistack.org/micro/v3 v3.11.37/go.mod h1:POGU5hstnAT9LH70m8FalyQSNi2GfIew71K75JenIZk=
+go.unistack.org/micro/v3 v3.11.45/go.mod h1:fDQ8Mu9wubaFP0L8hNQlpzHiEnWN0wbOlawN9HYo0N4=
-golang.org/x/crypto v0.31.0 h1:ihbySMvVjLAeSH1IbfcRTkD/iNscyz8rGzjF/E5hV6U=
+go.unistack.org/micro/v3 v3.11.48 h1:lHJYSHU2z1TTcuswItGwG7cZXN6n04EFqY7lk/0gA7w=
-golang.org/x/crypto v0.31.0/go.mod h1:kDsLvtWBEx7MV9tJOj9bnXsPbxwJQ6csT/x4KIN4Ssk=
+go.unistack.org/micro/v3 v3.11.48/go.mod h1:fDQ8Mu9wubaFP0L8hNQlpzHiEnWN0wbOlawN9HYo0N4=
-golang.org/x/net v0.33.0 h1:74SYHlV8BIgHIFC/LrYkOGIwL19eTYXQ5wc6TBuO36I=
+golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/net v0.33.0/go.mod h1:HXLR5J+9DxmrqMwG9qjGCxZ+zKXxBru04zlTvWlWuN4=
+golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
-golang.org/x/sys v0.28.0 h1:Fksou7UEQUWlKvIdsqzJmUmCX3cZuD2+P3XyyzwMhlA=
+golang.org/x/net v0.45.0 h1:RLBg5JKixCy82FtLJpeNlVM0nrSqpCRYzVU1n8kj0tM=
-golang.org/x/sys v0.28.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/net v0.45.0/go.mod h1:ECOoLqd5U3Lhyeyo/QDCEVQ4sNgYsqvCZ722XogGieY=
-golang.org/x/text v0.21.0 h1:zyQAAkrwaneQ066sspRyJaG9VNi/YJ1NfzcGB3hZ/qo=
+golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
-golang.org/x/text v0.21.0/go.mod h1:4IBbMaMmOPCJ8SecivzSH54+73PCFmPWxNTLm+vZkEQ=
+golang.org/x/sys v0.37.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20241216192217-9240e9c98484 h1:Z7FRVJPSMaHQxD0uXU8WdgFh8PseLM8Q8NzhnpMrBhQ=
+golang.org/x/text v0.30.0 h1:yznKA/E9zq54KzlzBEAWn1NXSQ8DIp/NYMy88xJjl4k=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20241216192217-9240e9c98484/go.mod h1:lcTa1sDdWEIHMWlITnIczmw5w60CF9ffkb8Z+DVmmjA=
+golang.org/x/text v0.30.0/go.mod h1:yDdHFIX9t+tORqspjENWgzaCVXgk0yYnYuSZ8UzzBVM=
-google.golang.org/grpc v1.69.2 h1:U3S9QEtbXC0bYNvRtcoklF3xGtLViumSYxWykJS+7AU=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251007200510-49b9836ed3ff h1:A90eA31Wq6HOMIQlLfzFwzqGKBTuaVztYu/g8sn+8Zc=
-google.golang.org/grpc v1.69.2/go.mod h1:vyjdE6jLBI76dgpDojsFGNaHlxdjXN9ghpnd2o7JGZ4=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251007200510-49b9836ed3ff/go.mod h1:7i2o+ce6H/6BluujYR+kqX3GKH+dChPTQU19wjRPiGk=
-google.golang.org/protobuf v1.36.1 h1:yBPeRvTftaleIgM3PZ/WBIZ7XM/eEYAaEyCwvyjq/gk=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251022142026-3a174f9686a8 h1:M1rk8KBnUsBDg1oPGHNCxG4vc1f49epmTO7xscSajMk=
-google.golang.org/protobuf v1.36.1/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251022142026-3a174f9686a8/go.mod h1:7i2o+ce6H/6BluujYR+kqX3GKH+dChPTQU19wjRPiGk=
+google.golang.org/grpc v1.76.0 h1:UnVkv1+uMLYXoIz6o7chp59WfQUYA2ex/BXQ9rHZu7A=
+google.golang.org/grpc v1.76.0/go.mod h1:Ju12QI8M6iQJtbcsV+awF5a4hfJMLi4X0JLo94ULZ6c=
+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=

hook_event.go

@@ -0,0 +1,73 @@
+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) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		m.log.Error(context.TODO(), "kgo.OnGroupManageError", err)
+	}
+}
+
+func (m *hookEvent) OnBrokerConnect(_ kgo.BrokerMetadata, _ time.Duration, _ net.Conn, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		m.log.Error(context.TODO(), "kgo.OnBrokerConnect", 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:
+		m.log.Error(context.TODO(), "kgo.OnBrokerWrite", 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:
+		m.log.Error(context.TODO(), "kgo.OnBrokerRead", err)
+	}
+}
+
+func (m *hookEvent) OnProduceRecordUnbuffered(_ *kgo.Record, err error) {
+	switch {
+	case err == nil || isContextError(err) || kgo.IsRetryableBrokerErr(err):
+		return
+	default:
+		m.log.Error(context.TODO(), "kgo.OnProduceRecordUnbuffered", err)
+	}
+}

hook_event_test.go

@@ -0,0 +1,362 @@
+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/v3/logger"
+)
+
+func TestHookEvent_OnGroupManageError(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "some error",
+			inputErr:              errors.New("some error"),
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnGroupManageError",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log}
+			he.OnGroupManageError(tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerConnect(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "some error",
+			inputErr:              errors.New("some error"),
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerConnect",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log}
+			he.OnBrokerConnect(kgo.BrokerMetadata{}, 0, nil, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerWrite(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "some error",
+			inputErr:              errors.New("some error"),
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerWrite",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log}
+			he.OnBrokerWrite(kgo.BrokerMetadata{}, 0, 0, 0, 0, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnBrokerRead(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "some error",
+			inputErr:              errors.New("some error"),
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnBrokerRead",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log}
+			he.OnBrokerRead(kgo.BrokerMetadata{}, 0, 0, 0, 0, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+		})
+	}
+}
+
+func TestHookEvent_OnProduceRecordUnbuffered(t *testing.T) {
+	tests := []struct {
+		name                  string
+		inputErr              error
+		expectedErrorIsCalled bool
+		expectedErrorMsg      string
+	}{
+		{
+			name:                  "error is nil",
+			inputErr:              nil,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context canceled",
+			inputErr:              context.Canceled,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "context deadline exceeded",
+			inputErr:              context.DeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: deadline exceeded (os package)",
+			inputErr:              os.ErrDeadlineExceeded,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: EOF (io package)",
+			inputErr:              io.EOF,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "retryable error: closed network connection (net package)",
+			inputErr:              net.ErrClosed,
+			expectedErrorIsCalled: false,
+		},
+		{
+			name:                  "some error",
+			inputErr:              errors.New("some error"),
+			expectedErrorIsCalled: true,
+			expectedErrorMsg:      "kgo.OnProduceRecordUnbuffered",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			log := &mockLogger{}
+			he := &hookEvent{log: log}
+			he.OnProduceRecordUnbuffered(&kgo.Record{}, tt.inputErr)
+			require.Equal(t, tt.expectedErrorIsCalled, log.errorIsCalled)
+			require.Equal(t, tt.expectedErrorMsg, log.errorMsg)
+		})
+	}
+}
+
+// Mocks
+
+type mockLogger struct {
+	errorIsCalled bool
+	errorMsg      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(context.Context, string, ...interface{}) {
+	panic("implement me")
+}
+
+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

@@ -6,13 +6,14 @@ import (
 	"errors"
 	"fmt"
 	"math/rand/v2"
-	"net/http"
+	"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"
@@ -20,6 +21,7 @@ import (
 	"go.unistack.org/micro/v3/metadata"
 	"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"
 )
 
@@ -60,12 +62,14 @@ type Broker struct {
 	c         *kgo.Client
 	connected *atomic.Uint32
 
+	done chan struct{}
+
 	kopts []kgo.Opt
 	subs  []*Subscriber
 
 	opts broker.Options
+	mu   sync.RWMutex
 
-	sync.RWMutex
 	init bool
 }
 
@@ -110,10 +114,18 @@ func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, *ho
 		}
 	}
 
+	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 {
@@ -135,8 +147,28 @@ func (k *Broker) connect(ctx context.Context, opts ...kgo.Opt) (*kgo.Client, *ho
 			}
 			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, htracer, nil
 }
 
 func (k *Broker) Connect(ctx context.Context) error {
@@ -154,10 +186,72 @@ func (k *Broker) Connect(ctx context.Context) error {
 		return err
 	}
 
-	k.Lock()
+	k.mu.Lock()
 	k.c = c
 	k.connected.Store(1)
-	k.Unlock()
+	k.mu.Unlock()
+
+	exposeLag := false
+	if k.opts.Context != nil {
+		if v, ok := k.opts.Context.Value(exposeLagKey{}).(bool); ok && v {
+			exposeLag = v
+		}
+	}
+
+	if exposeLag {
+		var mu sync.Mutex
+		var lastUpdate time.Time
+		type pl struct {
+			p string
+			l float64
+		}
+
+		lag := make(map[string]map[string]pl) // topic => group => partition => lag
+		ac := kadm.NewClient(k.c)
+
+		updateStats := func() {
+			mu.Lock()
+			if time.Since(lastUpdate) < DefaultStatsInterval {
+				return
+			}
+			mu.Unlock()
+
+			k.mu.Lock()
+			groups := make([]string, 0, len(k.subs))
+			for _, g := range k.subs {
+				groups = append(groups, g.opts.Group)
+			}
+			k.mu.Unlock()
+
+			dgls, err := ac.Lag(ctx, groups...)
+			if err != nil || !dgls.Ok() {
+				k.opts.Logger.Error(k.opts.Context, "kgo describe group lag error", err)
+				return
+			}
+
+			for gn, dgl := range dgls {
+				for tn, lmap := range dgl.Lag {
+					if _, ok := lag[tn]; !ok {
+						lag[tn] = make(map[string]pl)
+					}
+					for p, l := range lmap {
+						lag[tn][gn] = pl{p: strconv.Itoa(int(p)), l: float64(l.Lag)}
+					}
+				}
+			}
+		}
+
+		for tn, dg := range lag {
+			for gn, gl := range dg {
+				k.opts.Meter.Gauge(semconv.BrokerGroupLag,
+					func() float64 { updateStats(); return gl.l },
+					"topic", tn,
+					"group", gn,
+					"partition", gl.p)
+			}
+		}
+
+	}
 
 	return nil
 }
@@ -175,8 +269,8 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 	ctx, span = k.opts.Tracer.Start(ctx, "Disconnect")
 	defer span.Finish()
 
-	k.Lock()
+	k.mu.Lock()
-	defer k.Unlock()
+	defer k.mu.Unlock()
 	select {
 	case <-nctx.Done():
 		return nctx.Err()
@@ -196,12 +290,13 @@ func (k *Broker) Disconnect(ctx context.Context) error {
 	}
 
 	k.connected.Store(0)
+	close(k.done)
 	return nil
 }
 
 func (k *Broker) Init(opts ...broker.Option) error {
-	k.Lock()
+	k.mu.Lock()
-	defer k.Unlock()
+	defer k.mu.Unlock()
 
 	if len(opts) == 0 && k.init {
 		return nil
@@ -249,18 +344,6 @@ 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.Lock()
-	if k.connected.Load() == 0 {
-		c, _, err := k.connect(ctx, k.kopts...)
-		if err != nil {
-			k.Unlock()
-			return err
-		}
-		k.c = c
-		k.connected.Store(1)
-	}
-	k.Unlock()
-
 	options := broker.NewPublishOptions(opts...)
 	records := make([]*kgo.Record, 0, len(msgs))
 	var errs []string
@@ -279,14 +362,14 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 
 	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 {
+			setHeaders(rec, msg.Header)
-				rec.Headers = append(rec.Headers, kgo.RecordHeader{Key: http.CanonicalHeaderKey(k), Value: []byte(v)})
-			}
 		} else {
 			rec.Value, err = k.opts.Codec.Marshal(msg)
 			if err != nil {
@@ -315,7 +398,9 @@ func (k *Broker) publish(ctx context.Context, msgs []*broker.Message, opts ...br
 		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())
@@ -374,13 +459,22 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 		}
 	}
 
+	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,
+		topic:        topic,
-		opts:      options,
+		opts:         options,
-		handler:   handler,
+		handler:      handler,
-		kopts:     k.opts,
+		kopts:        k.opts,
-		consumers: make(map[tp]*consumer),
+		consumers:    make(map[tp]*consumer),
-		done:      make(chan struct{}),
+		done:         make(chan struct{}),
+		fatalOnError: fatalOnError,
+		connected:    k.connected,
 	}
 
 	kopts := append(k.kopts,
@@ -391,7 +485,9 @@ 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(),
 	)
 
@@ -423,9 +519,9 @@ func (k *Broker) Subscribe(ctx context.Context, topic string, handler broker.Han
 
 	go sub.poll(ctx)
 
-	k.Lock()
+	k.mu.Lock()
 	k.subs = append(k.subs, sub)
-	k.Unlock()
+	k.mu.Unlock()
 	return sub, nil
 }
 
@@ -452,7 +548,7 @@ func NewBroker(opts ...broker.Option) *Broker {
 		kgo.BlockRebalanceOnPoll(),
 		kgo.Balancers(kgo.CooperativeStickyBalancer()),
 		kgo.FetchIsolationLevel(kgo.ReadUncommitted()),
-		kgo.UnknownTopicRetries(0),
+		kgo.UnknownTopicRetries(1),
 	}
 
 	if options.Context != nil {
@@ -465,5 +561,6 @@ func NewBroker(opts ...broker.Option) *Broker {
 		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
+	prefill  = true
-	loglevel = logger.InfoLevel
+	loglevel = logger.DebugLevel
 )
 
 var bm = &broker.Message{
@@ -28,7 +28,99 @@ 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) {
+	if tr := os.Getenv("INTEGRATION_TESTS"); len(tr) > 0 {
+		t.Skip()
+	}
+
 	var addrs []string
 	ctx := context.TODO()
 	b := kgo.NewBroker(
@@ -67,6 +159,7 @@ 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 +217,7 @@ 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)

meter.go

@@ -32,27 +32,27 @@ var (
 )
 
 const (
-	metricBrokerConnects    = "broker_connects_total"
+	metricBrokerConnects    = "micro_broker_connects_total"
-	metricBrokerDisconnects = "broker_disconnects_total"
+	metricBrokerDisconnects = "micro_broker_disconnects_total"
 
-	metricBrokerWriteErrors        = "broker_write_errors_total"
+	metricBrokerWriteErrors        = "micro_broker_write_errors_total"
-	metricBrokerWriteBytes         = "broker_write_bytes_total"
+	metricBrokerWriteBytes         = "micro_broker_write_bytes_total"
-	metricBrokerWriteWaitLatencies = "broker_write_wait_latencies"
+	metricBrokerWriteWaitLatencies = "micro_broker_write_wait_latencies"
-	metricBrokerWriteLatencies     = "broker_write_latencies"
+	metricBrokerWriteLatencies     = "micro_broker_write_latencies"
 
-	metricBrokerReadErrors        = "broker_read_errors_total"
+	metricBrokerReadErrors        = "micro_broker_read_errors_total"
-	metricBrokerReadBytes         = "broker_read_bytes_total"
+	metricBrokerReadBytes         = "micro_broker_read_bytes_total"
-	metricBrokerReadWaitLatencies = "broker_read_wait_latencies"
+	metricBrokerReadWaitLatencies = "micro_broker_read_wait_latencies"
-	metricBrokerReadLatencies     = "broker_read_latencies"
+	metricBrokerReadLatencies     = "micro_broker_read_latencies"
 
-	metricBrokerThrottleLatencies = "broker_throttle_latencies"
+	metricBrokerThrottleLatencies = "micro_broker_throttle_latencies"
 
-	metricBrokerProduceBytesCompressed   = "broker_produce_bytes_compressed_total"
+	metricBrokerProduceBytesCompressed   = "micro_broker_produce_bytes_compressed_total"
-	metricBrokerProduceBytesUncompressed = "broker_produce_bytes_uncompressed_total"
+	metricBrokerProduceBytesUncompressed = "micro_broker_produce_bytes_uncompressed_total"
-	metricBrokerFetchBytesCompressed     = "broker_consume_bytes_compressed_total"
+	metricBrokerFetchBytesCompressed     = "micro_broker_consume_bytes_compressed_total"
-	metricBrokerFetchBytesUncompressed   = "broker_consume_bytes_uncompressed_total"
+	metricBrokerFetchBytesUncompressed   = "micro_broker_consume_bytes_uncompressed_total"
 
-	metricBrokerGroupErrors = "broker_group_errors_total"
+	metricBrokerGroupErrors = "micro_broker_group_errors_total"
 
 	labelNode    = "node_id"
 	labelSuccess = "success"

options.go

@@ -15,7 +15,7 @@ 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
@@ -72,6 +72,12 @@ 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 {
@@ -98,6 +104,11 @@ 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
@@ -109,3 +120,10 @@ func PublishPromise(fn func(*kgo.Record, error)) broker.PublishOption {
 func ClientPublishPromise(fn func(*kgo.Record, error)) client.PublishOption {
 	return client.SetPublishOption(publishPromiseKey{}, fn)
 }
+
+type exposeLagKey struct{}
+
+// ExposeLag enabled subscriber lag via [meter.Meter]
+func ExposeLag(b bool) broker.Option {
+	return broker.SetOption(exposeLagKey{}, b)
+}

subscriber.go

@@ -5,10 +5,11 @@ import (
 	"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"
@@ -24,13 +25,15 @@ type tp struct {
 type consumer struct {
 	topic string
 
-	c *kgo.Client
+	c         *kgo.Client
 	htracer   *hookTracer
+	connected *atomic.Uint32
+
+	quit chan struct{}
+	done chan struct{}
+	recs chan kgo.FetchTopicPartition
 
 	handler broker.Handler
-	quit    chan struct{}
-	done    chan struct{}
-	recs    chan kgo.FetchTopicPartition
 
 	kopts broker.Options
 	opts  broker.SubscribeOptions
@@ -39,18 +42,25 @@ type consumer struct {
 }
 
 type Subscriber struct {
+	topic string
+
 	consumers map[tp]*consumer
+
 	c         *kgo.Client
 	htracer   *hookTracer
-	topic     string
+	connected *atomic.Uint32
 
 	handler broker.Handler
-	done    chan struct{}
+
-	kopts   broker.Options
+	done chan struct{}
-	opts    broker.SubscribeOptions
+
+	kopts broker.Options
+	opts  broker.SubscribeOptions
+
+	closed       bool
+	fatalOnError bool
 
 	sync.RWMutex
-	closed bool
 }
 
 func (s *Subscriber) Client() *kgo.Client {
@@ -86,46 +96,13 @@ func (s *Subscriber) Unsubscribe(ctx context.Context) error {
 
 func (s *Subscriber) poll(ctx context.Context) {
 	maxInflight := DefaultSubscribeMaxInflight
+
 	if s.opts.Context != nil {
 		if n, ok := s.opts.Context.Value(subscribeMaxInflightKey{}).(int); n > 0 && ok {
 			maxInflight = n
 		}
 	}
 
-	go func() {
-		ac := kadm.NewClient(s.c)
-		ticker := time.NewTicker(DefaultStatsInterval)
-
-		for {
-			select {
-			case <-ctx.Done():
-				ticker.Stop()
-				return
-			case <-ticker.C:
-				dgls, err := ac.Lag(ctx, s.opts.Group)
-				if err != nil || !dgls.Ok() {
-					continue
-				}
-
-				dgl, ok := dgls[s.opts.Group]
-				if !ok {
-					continue
-				}
-				lmap, ok := dgl.Lag[s.topic]
-				if !ok {
-					continue
-				}
-
-				s.Lock()
-				for p, l := range lmap {
-					s.kopts.Meter.Counter(semconv.BrokerGroupLag, "topic", s.topic, "group", s.opts.Group, "partition", strconv.Itoa(int(p)), "lag", strconv.Itoa(int(l.Lag)))
-				}
-				s.Unlock()
-
-			}
-		}
-	}()
-
 	for {
 		select {
 		case <-ctx.Done():
@@ -140,12 +117,16 @@ func (s *Subscriber) poll(ctx context.Context) {
 				return
 			}
 			fetches.EachError(func(t string, p int32, err error) {
-				s.kopts.Logger.Fatal(ctx, fmt.Sprintf("[kgo] fetch topic %s partition %d error", 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) {
-				nTp := tp{p.Topic, p.Partition}
+				tps := tp{p.Topic, p.Partition}
-				s.consumers[nTp].recs <- p
+				s.consumers[tps].recs <- p
 			})
 			s.c.AllowRebalance()
 		}
@@ -158,9 +139,12 @@ func (s *Subscriber) killConsumers(ctx context.Context, lost map[string][]int32)
 
 	for topic, partitions := range lost {
 		for _, partition := range partitions {
-			nTp := tp{topic, partition}
+			tps := tp{topic, partition}
-			pc := s.consumers[nTp]
+			pc, ok := s.consumers[tps]
-			delete(s.consumers, nTp)
+			if !ok {
+				continue
+			}
+			delete(s.consumers, tps)
 			close(pc.quit)
 			if s.kopts.Logger.V(logger.DebugLevel) {
 				s.kopts.Logger.Debug(ctx, fmt.Sprintf("[kgo] waiting for work to finish topic %s partition %d", topic, partition))
@@ -171,11 +155,21 @@ func (s *Subscriber) killConsumers(ctx context.Context, lost map[string][]int32)
 	}
 }
 
+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) lost(ctx context.Context, _ *kgo.Client, lost map[string][]int32) {
-	if s.kopts.Logger.V(logger.DebugLevel) {
+	if s.kopts.Logger.V(logger.ErrorLevel) {
-		s.kopts.Logger.Debug(ctx, fmt.Sprintf("[kgo] lost %#+v", lost))
+		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) {
@@ -185,6 +179,7 @@ func (s *Subscriber) revoked(ctx context.Context, c *kgo.Client, revoked map[str
 	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)
 	}
 }
 
@@ -202,6 +197,7 @@ func (s *Subscriber) assigned(_ context.Context, c *kgo.Client, assigned map[str
 				handler:   s.handler,
 				kopts:     s.kopts,
 				opts:      s.opts,
+				connected: s.connected,
 			}
 			s.Lock()
 			s.consumers[tp{topic, partition}] = pc
@@ -243,16 +239,27 @@ func (pc *consumer) consume() {
 				for _, hdr := range record.Headers {
 					p.msg.Header.Set(hdr.Key, string(hdr.Value))
 				}
+				p.msg.Header.Set("Micro-Offset", strconv.FormatInt(record.Offset, 10))
+				p.msg.Header.Set("Micro-Partition", strconv.FormatInt(int64(record.Partition), 10))
+				p.msg.Header.Set("Micro-Topic", record.Topic)
+				p.msg.Header.Set("Micro-Key", string(record.Key))
+				p.msg.Header.Set("Micro-Timestamp", strconv.FormatInt(record.Timestamp.Unix(), 10))
 				if pc.kopts.Codec.String() == "noop" {
 					p.msg.Body = record.Value
 				} else if pc.opts.BodyOnly {
 					p.msg.Body = record.Value
 				} else {
-					sp.AddEvent("codec unmarshal start")
+					if sp != nil {
+						sp.AddEvent("codec unmarshal start")
+					}
 					err := pc.kopts.Codec.Unmarshal(record.Value, p.msg)
-					sp.AddEvent("codec unmarshal stop")
+					if sp != nil {
+						sp.AddEvent("codec unmarshal stop")
+					}
 					if err != nil {
-						sp.SetStatus(tracer.SpanStatusError, err.Error())
+						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
@@ -263,6 +270,7 @@ func (pc *consumer) consume() {
 								pc.c.MarkCommitRecords(record)
 							} else {
 								eventPool.Put(p)
+								//								pc.connected.Store(0)
 								pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
 								return
 							}
@@ -279,18 +287,27 @@ func (pc *consumer) consume() {
 						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.connected.Store(0)
 						pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] Unmarshal err not handled wtf?")
-						sp.Finish()
+						if sp != nil {
+							sp.Finish()
+						}
 						return
 					}
 				}
-				sp.AddEvent("handler start")
+				if sp != nil {
+					sp.AddEvent("handler start")
+				}
 				err := pc.handler(p)
-				sp.AddEvent("handler stop")
+				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 {
-					sp.SetStatus(tracer.SpanStatusError, err.Error())
+					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()
@@ -299,9 +316,13 @@ func (pc *consumer) consume() {
 				} else if err != nil {
 					p.err = err
 					if eh != nil {
-						sp.AddEvent("error handler start")
+						if sp != nil {
+							sp.AddEvent("error handler start")
+						}
 						_ = eh(p)
-						sp.AddEvent("error handler stop")
+						if sp != nil {
+							sp.AddEvent("error handler stop")
+						}
 					} else {
 						if pc.kopts.Logger.V(logger.ErrorLevel) {
 							pc.kopts.Logger.Error(pc.kopts.Context, "[kgo]: subscriber error", err)
@@ -316,12 +337,17 @@ func (pc *consumer) consume() {
 					pc.c.MarkCommitRecords(record)
 				} else {
 					eventPool.Put(p)
+					//					pc.connected.Store(0)
 					pc.kopts.Logger.Fatal(pc.kopts.Context, "[kgo] ErrLostMessage wtf?")
-					sp.SetStatus(tracer.SpanStatusError, "ErrLostMessage")
+					if sp != nil {
-					sp.Finish()
+						sp.SetStatus(tracer.SpanStatusError, "ErrLostMessage")
+						sp.Finish()
+					}
 					return
 				}
-				sp.Finish()
+				if sp != nil {
+					sp.Finish()
+				}
 			}
 		}
 	}

tracer.go

@@ -32,6 +32,9 @@ var (
 // the record's context, so it can be ended in the OnProduceRecordUnbuffered
 // hook.
 func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	// Set up span options.
 	attrs := []interface{}{
 		messagingSystem,
@@ -68,7 +71,7 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
 	}
 
-	setHeaders(r, omd)
+	setHeaders(r, omd, metadata.HeaderContentType)
 }
 
 // OnProduceRecordUnbuffered continues and ends the "publish" span for an
@@ -77,6 +80,9 @@ func (m *hookTracer) OnProduceRecordBuffered(r *kgo.Record) {
 // It sets attributes with values unset when producing and records any error
 // that occurred during the publish operation.
 func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	if span, ok := tracer.SpanFromContext(r.Context); ok {
 		span.AddLabels(
 			semconv.MessagingKafkaDestinationPartition(int(r.Partition)),
@@ -96,6 +102,9 @@ func (m *hookTracer) OnProduceRecordUnbuffered(r *kgo.Record, err error) {
 // OnFetchRecordUnbuffered hook and can be used in downstream consumer
 // processing.
 func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
+	if !m.tracer.Enabled() {
+		return
+	}
 	// Set up the span options.
 	attrs := []interface{}{
 		messagingSystem,
@@ -135,12 +144,15 @@ func (m *hookTracer) OnFetchRecordBuffered(r *kgo.Record) {
 		r.Context, _ = m.tracer.Start(r.Context, "sdk.broker", opts...)
 	}
 
-	setHeaders(r, omd)
+	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()
 }
@@ -155,6 +167,13 @@ func (m *hookTracer) OnFetchRecordUnbuffered(r *kgo.Record, _ bool) {
 // not a record which has been created for producing, so call this at the start of each
 // iteration of your processing for the record.
 func (m *hookTracer) WithProcessSpan(r *kgo.Record) (context.Context, tracer.Span) {
+	if r.Context == nil {
+		r.Context = context.Background()
+	}
+
+	if !m.tracer.Enabled() {
+		return r.Context, nil
+	}
 	// Set up the span options.
 	attrs := []interface{}{
 		messagingSystem,