Merge branch 'main' into mcarl/lint

Signed-off-by: Julien <291750+roidelapluie@users.noreply.github.com>

.github/CODEOWNERS

@@ -1,7 +1,7 @@
 /web/ui @juliusv
 /web/ui/module @juliusv @nexucis
-/storage/remote @csmarchbanks @cstyan @bwplotka @tomwilkie
-/storage/remote/otlptranslator @gouthamve @jesusvazquez
+/storage/remote @cstyan @bwplotka @tomwilkie
+/storage/remote/otlptranslator @aknuds1 @jesusvazquez
 /discovery/kubernetes @brancz
 /tsdb @jesusvazquez
 /promql @roidelapluie

.github/dependabot.yml

@@ -6,11 +6,11 @@ updates:
       interval: "monthly"
     groups:
       k8s.io:
-          patterns:
-            - "k8s.io/*"
+        patterns:
+          - "k8s.io/*"
       go.opentelemetry.io:
-          patterns:
-            - "go.opentelemetry.io/*"
+        patterns:
+          - "go.opentelemetry.io/*"
   - package-ecosystem: "gomod"
     directory: "/documentation/examples/remote_storage"
     schedule:

.github/workflows/buf-lint.yml

@@ -12,14 +12,14 @@ jobs:
     name: lint
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
-      - uses: bufbuild/buf-setup-action@382440cdb8ec7bc25a68d7b4711163d95f7cc3aa # v1.28.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - uses: bufbuild/buf-setup-action@517ee23296d5caf38df31c21945e6a54bbc8a89f # v1.30.0
         with:
           github_token: ${{ secrets.GITHUB_TOKEN }}
-      - uses: bufbuild/buf-lint-action@bd48f53224baaaf0fc55de9a913e7680ca6dbea4 # v1.0.3
+      - uses: bufbuild/buf-lint-action@06f9dd823d873146471cfaaf108a993fe00e5325 # v1.1.1
         with:
           input: 'prompb'
-      - uses: bufbuild/buf-breaking-action@f47418c81c00bfd65394628385593542f64db477 # v1.1.2
+      - uses: bufbuild/buf-breaking-action@c57b3d842a5c3f3b454756ef65305a50a587c5ba # v1.1.4
         with:
           input: 'prompb'
           against: 'https://github.com/prometheus/prometheus.git#branch=main,ref=HEAD,subdir=prompb'

.github/workflows/buf.yml

@@ -12,18 +12,18 @@ jobs:
     runs-on: ubuntu-latest
     if: github.repository_owner == 'prometheus'
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
-      - uses: bufbuild/buf-setup-action@382440cdb8ec7bc25a68d7b4711163d95f7cc3aa # v1.28.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - uses: bufbuild/buf-setup-action@517ee23296d5caf38df31c21945e6a54bbc8a89f # v1.30.0
         with:
           github_token: ${{ secrets.GITHUB_TOKEN }}
-      - uses: bufbuild/buf-lint-action@bd48f53224baaaf0fc55de9a913e7680ca6dbea4 # v1.0.3
+      - uses: bufbuild/buf-lint-action@06f9dd823d873146471cfaaf108a993fe00e5325 # v1.1.1
         with:
           input: 'prompb'
-      - uses: bufbuild/buf-breaking-action@f47418c81c00bfd65394628385593542f64db477 # v1.1.2
+      - uses: bufbuild/buf-breaking-action@c57b3d842a5c3f3b454756ef65305a50a587c5ba # v1.1.4
         with:
           input: 'prompb'
           against: 'https://github.com/prometheus/prometheus.git#branch=main,ref=HEAD~1,subdir=prompb'
-      - uses: bufbuild/buf-push-action@342fc4cdcf29115a01cf12a2c6dd6aac68dc51e1 # v1.1.1
+      - uses: bufbuild/buf-push-action@a654ff18effe4641ebea4a4ce242c49800728459 # v1.1.1
         with:
           input: 'prompb'
           buf_token: ${{ secrets.BUF_TOKEN }}

.github/workflows/ci.yml

@@ -8,34 +8,56 @@ jobs:
   test_go:
     name: Go tests
     runs-on: ubuntu-latest
-    # Whenever the Go version is updated here, .promu.yml
-    # should also be updated.
     container:
-      image: quay.io/prometheus/golang-builder:1.21-base
+      # Whenever the Go version is updated here, .promu.yml
+      # should also be updated.
+      image: quay.io/prometheus/golang-builder:1.22-base
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/setup_environment
-      - run: make GO_ONLY=1 SKIP_GOLANGCI_LINT=1
-      - run: go test ./tsdb/ -test.tsdb-isolation=false
-      - run: go test --tags=stringlabels ./...
-      - run: GOARCH=386 go test ./cmd/prometheus
+      - run: make GOOPTS=--tags=stringlabels GO_ONLY=1 SKIP_GOLANGCI_LINT=1
+      - run: go test --tags=stringlabels ./tsdb/ -test.tsdb-isolation=false
       - run: make -C documentation/examples/remote_storage
       - run: make -C documentation/examples
+
+  test_go_more:
+    name: More Go tests
+    runs-on: ubuntu-latest
+    container:
+      image: quay.io/prometheus/golang-builder:1.22-base
+    steps:
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
+      - uses: ./.github/promci/actions/setup_environment
+      - run: go test --tags=dedupelabels ./...
+      - run: GOARCH=386 go test ./cmd/prometheus
       - uses: ./.github/promci/actions/check_proto
         with:
           version: "3.15.8"
 
+  test_go_oldest:
+    name: Go tests with previous Go version
+    runs-on: ubuntu-latest
+    container:
+      # The go version in this image should be N-1 wrt test_go.
+      image: quay.io/prometheus/golang-builder:1.21-base
+    steps:
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - run: make build
+      # Don't run NPM build; don't run race-detector.
+      - run: make test GO_ONLY=1 test-flags=""
+
   test_ui:
     name: UI tests
     runs-on: ubuntu-latest
     # Whenever the Go version is updated here, .promu.yml
     # should also be updated.
     container:
-      image: quay.io/prometheus/golang-builder:1.21-base
+      image: quay.io/prometheus/golang-builder:1.22-base
 
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/setup_environment
         with:
@@ -52,36 +74,24 @@ jobs:
     name: Go tests on Windows
     runs-on: windows-latest
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
-      - uses: actions/setup-go@93397bea11091df50f3d7e59dc26a7711a8bcfbe # v4.1.0
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
         with:
-          go-version: 1.21.x
+          go-version: 1.22.x
       - run: |
           $TestTargets = go list ./... | Where-Object { $_ -NotMatch "(github.com/prometheus/prometheus/discovery.*|github.com/prometheus/prometheus/config|github.com/prometheus/prometheus/web)"}
           go test $TestTargets -vet=off -v
         shell: powershell
 
-  test_golang_oldest:
-    name: Go tests with previous Go version
-    runs-on: ubuntu-latest
-    # The go verson in this image should be N-1 wrt test_go.
-    container:
-      image: quay.io/prometheus/golang-builder:1.20-base
-    steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
-      - run: make build
-      - run: go test ./tsdb/...
-      - run: go test ./tsdb/ -test.tsdb-isolation=false
-
   test_mixins:
     name: Mixins tests
     runs-on: ubuntu-latest
     # Whenever the Go version is updated here, .promu.yml
     # should also be updated.
     container:
-      image: quay.io/prometheus/golang-builder:1.20-base
+      image: quay.io/prometheus/golang-builder:1.22-base
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - run: go install ./cmd/promtool/.
       - run: go install github.com/google/go-jsonnet/cmd/jsonnet@latest
       - run: go install github.com/google/go-jsonnet/cmd/jsonnetfmt@latest
@@ -104,7 +114,7 @@ jobs:
       matrix:
         thread: [ 0, 1, 2 ]
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/build
         with:
@@ -127,32 +137,45 @@ jobs:
     # Whenever the Go version is updated here, .promu.yml
     # should also be updated.
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/build
         with:
           parallelism: 12
           thread: ${{ matrix.thread }}
+  check_generated_parser:
+    name: Check generated parser
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - name: Install Go
+        uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
+        with:
+          cache: false
+          go-version: 1.22.x
+      - name: Run goyacc and check for diff
+        run: make install-goyacc check-generated-parser
   golangci:
     name: golangci-lint
     runs-on: ubuntu-latest
     steps:
       - name: Checkout repository
-        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - name: Install Go
-        uses: actions/setup-go@93397bea11091df50f3d7e59dc26a7711a8bcfbe # v4.1.0
+        uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0
         with:
           cache: false
-          go-version: 1.21.x
+          go-version: 1.22.x
       - name: Install snmp_exporter/generator dependencies
         run: sudo apt-get update && sudo apt-get -y install libsnmp-dev
         if: github.repository == 'prometheus/snmp_exporter'
       - name: Lint
-        uses: golangci/golangci-lint-action@3a919529898de77ec3da873e3063ca4b10e7f5cc # v3.7.0
+        uses: golangci/golangci-lint-action@3cfe3a4abbb849e10058ce4af15d205b6da42804 # v4.0.0
         with:
           args: --verbose
           # Make sure to sync this with Makefile.common and scripts/golangci-lint.yml.
-          version: v1.55.2
+          version: v1.56.2
   fuzzing:
     uses: ./.github/workflows/fuzzing.yml
     if: github.event_name == 'pull_request'
@@ -162,10 +185,10 @@ jobs:
   publish_main:
     name: Publish main branch artifacts
     runs-on: ubuntu-latest
-    needs: [test_ui, test_go, test_windows, golangci, codeql, build_all]
+    needs: [test_ui, test_go, test_go_more, test_go_oldest, test_windows, golangci, codeql, build_all]
     if: github.event_name == 'push' && github.event.ref == 'refs/heads/main'
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/publish_main
         with:
@@ -176,10 +199,10 @@ jobs:
   publish_release:
     name: Publish release artefacts
     runs-on: ubuntu-latest
-    needs: [test_ui, test_go, test_windows, golangci, codeql, build_all]
+    needs: [test_ui, test_go, test_go_more, test_go_oldest, test_windows, golangci, codeql, build_all]
     if: github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v2.')
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - uses: ./.github/promci/actions/publish_release
         with:
@@ -194,14 +217,14 @@ jobs:
     needs: [test_ui, codeql]
     steps:
       - name: Checkout
-        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - uses: prometheus/promci@3cb0c3871f223bd5ce1226995bd52ffb314798b6 # v0.1.0
       - name: Install nodejs
-        uses: actions/setup-node@5e21ff4d9bc1a8cf6de233a3057d20ec6b3fb69d # v3.8.1
+        uses: actions/setup-node@60edb5dd545a775178f52524783378180af0d1f8 # v4.0.2
         with:
           node-version-file: "web/ui/.nvmrc"
           registry-url: "https://registry.npmjs.org"
-      - uses: actions/cache@88522ab9f39a2ea568f7027eddc7d8d8bc9d59c8 # v3.3.1
+      - uses: actions/cache@0c45773b623bea8c8e75f6c82b208c3cf94ea4f9 # v4.0.2
         with:
           path: ~/.npm
           key: ${{ runner.os }}-node-${{ hashFiles('**/package-lock.json') }}

.github/workflows/codeql-analysis.yml

@@ -20,22 +20,19 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        language: ["go", "javascript"]
+        language: ["javascript"]
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
-      - uses: actions/setup-go@93397bea11091df50f3d7e59dc26a7711a8bcfbe # v4.1.0
-        with:
-          go-version: 1.21.x
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
 
       - name: Initialize CodeQL
-        uses: github/codeql-action/init@407ffafae6a767df3e0230c3df91b6443ae8df75 # v2.22.8
+        uses: github/codeql-action/init@012739e5082ff0c22ca6d6ab32e07c36df03c4a4 # v3.22.12
         with:
           languages: ${{ matrix.language }}
 
       - name: Autobuild
-        uses: github/codeql-action/autobuild@407ffafae6a767df3e0230c3df91b6443ae8df75 # v2.22.8
+        uses: github/codeql-action/autobuild@012739e5082ff0c22ca6d6ab32e07c36df03c4a4 # v3.22.12
 
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@407ffafae6a767df3e0230c3df91b6443ae8df75 # v2.22.8
+        uses: github/codeql-action/analyze@012739e5082ff0c22ca6d6ab32e07c36df03c4a4 # v3.22.12

.github/workflows/container_description.yml

@@ -0,0 +1,52 @@
+---
+name: Push README to Docker Hub
+on:
+  push:
+    paths:
+      - "README.md"
+      - ".github/workflows/container_description.yml"
+    branches: [ main, master ]
+
+permissions:
+  contents: read
+
+jobs:
+  PushDockerHubReadme:
+    runs-on: ubuntu-latest
+    name: Push README to Docker Hub
+    if: github.repository_owner == 'prometheus' || github.repository_owner == 'prometheus-community' # Don't run this workflow on forks.
+    steps:
+      - name: git checkout
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - name: Set docker hub repo name
+        run: echo "DOCKER_REPO_NAME=$(make docker-repo-name)" >> $GITHUB_ENV
+      - name: Push README to Dockerhub
+        uses: christian-korneck/update-container-description-action@d36005551adeaba9698d8d67a296bd16fa91f8e8 # v1
+        env:
+          DOCKER_USER: ${{ secrets.DOCKER_HUB_LOGIN }}
+          DOCKER_PASS: ${{ secrets.DOCKER_HUB_PASSWORD }}
+        with:
+          destination_container_repo: ${{ env.DOCKER_REPO_NAME }}
+          provider: dockerhub
+          short_description: ${{ env.DOCKER_REPO_NAME }}
+          readme_file: 'README.md'
+
+  PushQuayIoReadme:
+    runs-on: ubuntu-latest
+    name: Push README to quay.io
+    if: github.repository_owner == 'prometheus' || github.repository_owner == 'prometheus-community' # Don't run this workflow on forks.
+    steps:
+      - name: git checkout
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
+      - name: Set quay.io org name
+        run: echo "DOCKER_REPO=$(echo quay.io/${GITHUB_REPOSITORY_OWNER} | tr -d '-')" >> $GITHUB_ENV
+      - name: Set quay.io repo name
+        run: echo "DOCKER_REPO_NAME=$(make docker-repo-name)" >> $GITHUB_ENV
+      - name: Push README to quay.io
+        uses: christian-korneck/update-container-description-action@d36005551adeaba9698d8d67a296bd16fa91f8e8 # v1
+        env:
+          DOCKER_APIKEY: ${{ secrets.QUAY_IO_API_TOKEN }}
+        with:
+          destination_container_repo: ${{ env.DOCKER_REPO_NAME }}
+          provider: quay
+          readme_file: 'README.md'

.github/workflows/fuzzing.yml

@@ -21,7 +21,7 @@ jobs:
           fuzz-seconds: 600
           dry-run: false
       - name: Upload Crash
-        uses: actions/upload-artifact@a8a3f3ad30e3422c9c7b888a15615d19a852ae32 # v3.1.3
+        uses: actions/upload-artifact@65462800fd760344b1a7b4382951275a0abb4808 # v4.3.3
         if: failure() && steps.build.outcome == 'success'
         with:
           name: artifacts

.github/workflows/repo_sync.yml

@@ -13,7 +13,7 @@ jobs:
     container:
       image: quay.io/prometheus/golang-builder
     steps:
-      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+      - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4
       - run: ./scripts/sync_repo_files.sh
         env:
           GITHUB_TOKEN: ${{ secrets.PROMBOT_GITHUB_TOKEN }}

.github/workflows/scorecards.yml

@@ -21,7 +21,7 @@ jobs:
 
     steps:
       - name: "Checkout code"
-        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # tag=v4.1.1
+        uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # tag=v4.1.4
         with:
           persist-credentials: false
 
@@ -37,7 +37,7 @@ jobs:
       # Upload the results as artifacts (optional). Commenting out will disable uploads of run results in SARIF
       # format to the repository Actions tab.
       - name: "Upload artifact"
-        uses: actions/upload-artifact@a8a3f3ad30e3422c9c7b888a15615d19a852ae32 # tag=v3.1.3
+        uses: actions/upload-artifact@65462800fd760344b1a7b4382951275a0abb4808 # tag=v4.3.3
         with:
           name: SARIF file
           path: results.sarif
@@ -45,6 +45,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@407ffafae6a767df3e0230c3df91b6443ae8df75 # tag=v2.22.8
+        uses: github/codeql-action/upload-sarif@012739e5082ff0c22ca6d6ab32e07c36df03c4a4 # tag=v3.22.12
         with:
           sarif_file: results.sarif

.golangci.yml

@@ -21,11 +21,14 @@ linters:
     - goimports
     - misspell
     - nolintlint
+    - perfsprint
     - predeclared
     - revive
     - testifylint
     - unconvert
     - unused
+    - usestdlibvars
+    - whitespace
 
 issues:
   max-same-issues: 0
@@ -42,24 +45,32 @@ issues:
     - linters:
         - godot
       source: "^// ==="
-
+    - linters:
+        - perfsprint
+      text: "fmt.Sprintf can be replaced with string addition"
 linters-settings:
   depguard:
     rules:
       main:
         deny:
-        - pkg: "sync/atomic"
-          desc: "Use go.uber.org/atomic instead of sync/atomic"
-        - pkg: "github.com/stretchr/testify/assert"
-          desc: "Use github.com/stretchr/testify/require instead of github.com/stretchr/testify/assert"
-        - pkg: "github.com/go-kit/kit/log"
-          desc: "Use github.com/go-kit/log instead of github.com/go-kit/kit/log"
-        - pkg: "io/ioutil"
-          desc: "Use corresponding 'os' or 'io' functions instead."
-        - pkg: "regexp"
-          desc: "Use github.com/grafana/regexp instead of regexp"
-        - pkg: "github.com/pkg/errors"
-          desc: "Use 'errors' or 'fmt' instead of github.com/pkg/errors"
+          - pkg: "sync/atomic"
+            desc: "Use go.uber.org/atomic instead of sync/atomic"
+          - pkg: "github.com/stretchr/testify/assert"
+            desc: "Use github.com/stretchr/testify/require instead of github.com/stretchr/testify/assert"
+          - pkg: "github.com/go-kit/kit/log"
+            desc: "Use github.com/go-kit/log instead of github.com/go-kit/kit/log"
+          - pkg: "io/ioutil"
+            desc: "Use corresponding 'os' or 'io' functions instead."
+          - pkg: "regexp"
+            desc: "Use github.com/grafana/regexp instead of regexp"
+          - pkg: "github.com/pkg/errors"
+            desc: "Use 'errors' or 'fmt' instead of github.com/pkg/errors"
+          - pkg: "gzip"
+            desc: "Use github.com/klauspost/compress instead of gzip"
+          - pkg: "zlib"
+            desc: "Use github.com/klauspost/compress instead of zlib"
+          - pkg: "golang.org/x/exp/slices"
+            desc: "Use 'slices' instead."
   errcheck:
     exclude-functions:
       # Don't flag lines such as "io.Copy(io.Discard, resp.Body)".
@@ -77,6 +88,9 @@ linters-settings:
     local-prefixes: github.com/prometheus/prometheus
   gofumpt:
     extra-rules: true
+  perfsprint:
+     # Optimizes `fmt.Errorf`.
+    errorf: false
   revive:
     # By default, revive will enable only the linting rules that are named in the configuration file.
     # So, it's needed to explicitly set in configuration all required rules.
@@ -129,4 +143,3 @@ linters-settings:
       - require-error
       - suite-dont-use-pkg
       - suite-extra-assert-call
- 

.promu.yml

@@ -1,7 +1,7 @@
 go:
     # Whenever the Go version is updated here,
-    # .circle/config.yml should also be updated.
-    version: 1.21
+    # .github/workflows should also be updated.
+    version: 1.22
 repository:
     path: github.com/prometheus/prometheus
 build:

.yamllint

@@ -1,5 +1,7 @@
 ---
 extends: default
+ignore: |
+  ui/react-app/node_modules
 
 rules:
   braces:

CHANGELOG.md

@@ -2,8 +2,164 @@
 
 ## unreleased
 
-* [ENHANCEMENT] TSDB: Make the wlog watcher read segments synchronously when not tailing. #13224
-* [BUGFIX] Agent: Participate in notify calls. #13223
+* [CHANGE] Rules: Execute 1 query instead of N (where N is the number of alerts within alert rule) when restoring alerts. #13980
+* [ENHANCEMENT] Rules: Add `rule_group_last_restore_duration_seconds` to measure the time it takes to restore a rule group. #13974
+* [ENHANCEMENT] OTLP: Improve remote write format translation performance by using label set hashes for metric identifiers instead of string based ones. #14006 #13991
+* [ENHANCEMENT] TSDB: Optimize querying with regexp matchers. #13620
+* [BUGFIX] OTLP: Don't generate target_info unless at least one identifying label is defined. #13991
+* [BUGFIX] OTLP: Don't generate target_info unless there are metrics. #13991
+
+## 2.52.0-rc.1 / 2024-05-03
+
+* [BUGFIX] API: Fix missing comma during JSON encoding of API results. #14047
+
+## 2.52.0-rc.0 / 2024-04-22
+
+* [CHANGE] TSDB: Fix the predicate checking for blocks which are beyond the retention period to include the ones right at the retention boundary. #9633
+* [FEATURE] Kubernetes SD: Add a new metric `prometheus_sd_kubernetes_failures_total` to track failed requests to Kubernetes API. #13554
+* [FEATURE] Kubernetes SD: Add node and zone metadata labels when using the endpointslice role. #13935
+* [FEATURE] Azure SD/Remote Write: Allow usage of Azure authorization SDK. #13099
+* [FEATURE] Alerting: Support native histogram templating. #13731
+* [FEATURE] Linode SD: Support IPv6 range discovery and region filtering. #13774
+* [ENHANCEMENT] PromQL: Performance improvements for queries with regex matchers. #13461
+* [ENHANCEMENT] PromQL: Performance improvements when using aggregation operators. #13744
+* [ENHANCEMENT] PromQL: Validate label_join destination label. #13803
+* [ENHANCEMENT] Scrape: Increment `prometheus_target_scrapes_sample_duplicate_timestamp_total` metric on duplicated series during one scrape. #12933
+* [ENHANCEMENT] TSDB: Many improvements in performance. #13742 #13673 #13782
+* [ENHANCEMENT] TSDB: Pause regular block compactions if the head needs to be compacted (prioritize head as it increases memory consumption). #13754
+* [ENHANCEMENT] Observability: Improved logging during signal handling termination. #13772
+* [ENHANCEMENT] Observability: All log lines for drop series use "num_dropped" key consistently. #13823
+* [ENHANCEMENT] Observability: Log chunk snapshot and mmaped chunk replay duration during WAL replay. #13838
+* [ENHANCEMENT] Observability: Log if the block is being created from WBL during compaction. #13846
+* [BUGFIX] PromQL: Fix inaccurate sample number statistic when querying histograms. #13667
+* [BUGFIX] PromQL: Fix `histogram_stddev` and `histogram_stdvar` for cases where the histogram has negative buckets. #13852
+* [BUGFIX] PromQL: Fix possible duplicated label name and values in a metric result for specific queries. #13845
+* [BUGFIX] Scrape: Fix setting native histogram schema factor during scrape. #13846
+* [BUGFIX] TSDB: Fix counting of histogram samples when creating WAL checkpoint stats. #13776
+* [BUGFIX] TSDB: Fix cases of compacting empty heads. #13755
+* [BUGFIX] TSDB: Count float histograms in WAL checkpoint. #13844
+* [BUGFIX] Remote Read: Fix memory leak due to broken requests. #13777
+* [BUGFIX] API: Stop building response for `/api/v1/series/` when the API request was cancelled. #13766
+* [BUGFIX] promtool: Fix panic on `promtool tsdb analyze --extended` when no native histograms are present. #13976
+
+## 2.51.2 / 2024-04-09
+
+Bugfix release.
+
+[BUGFIX] Notifier: could hang when using relabeling on alerts #13861
+
+## 2.51.1 / 2024-03-27
+
+Bugfix release.
+
+* [BUGFIX] PromQL: Re-instate validation of label_join destination label #13803
+* [BUGFIX] Scraping (experimental native histograms): Fix handling of the min bucket factor on sync of targets #13846
+* [BUGFIX] PromQL: Some queries could return the same series twice (library use only) #13845
+
+## 2.51.0 / 2024-03-18
+
+This version is built with Go 1.22.1.
+
+There is a new optional build tag "dedupelabels", which should reduce memory consumption (#12304).
+It is off by default; there will be an optional alternative image to try it out.
+
+* [CHANGE] Scraping: Do experimental timestamp alignment even if tolerance is bigger than 1% of scrape interval #13624, #13737
+* [FEATURE] Alerting: Relabel rules for AlertManagerConfig; allows routing alerts to different alertmanagers #12551, #13735
+* [FEATURE] API: add limit param to series, label-names and label-values APIs #13396
+* [FEATURE] UI (experimental native histograms): Add native histogram chart to Table view #13658
+* [FEATURE] Promtool: Add a "tsdb dump-openmetrics" to dump in OpenMetrics format. #13194
+* [FEATURE] PromQL (experimental native histograms): Add histogram_avg function #13467
+* [ENHANCEMENT] Rules: Evaluate independent rules concurrently #12946, #13527
+* [ENHANCEMENT] Scraping (experimental native histograms): Support exemplars #13488
+* [ENHANCEMENT] Remote Write: Disable resharding during active retry backoffs #13562
+* [ENHANCEMENT] Observability: Add native histograms to latency/duration metrics #13681
+* [ENHANCEMENT] Observability: Add 'type' label to prometheus_tsdb_head_out_of_order_samples_appended_total #13607
+* [ENHANCEMENT] API: Faster generation of targets into JSON #13469, #13484
+* [ENHANCEMENT] Scraping, API: Use faster compression library #10782
+* [ENHANCEMENT] OpenTelemetry: Performance improvements in OTLP parsing #13627
+* [ENHANCEMENT] PromQL: Optimisations to reduce CPU and memory #13448, #13536
+* [BUGFIX] PromQL: Constrain extrapolation in rate() to half of sample interval #13725
+* [BUGFIX] Remote Write: Stop slowing down when a new WAL segment is created #13583, #13628
+* [BUGFIX] PromQL: Fix wrongly scoped range vectors with @ modifier #13559
+* [BUGFIX] Kubernetes SD: Pod status changes were not discovered by Endpoints service discovery #13337
+* [BUGFIX] Azure SD: Fix 'error: parameter virtualMachineScaleSetName cannot be empty' (#13702)
+* [BUGFIX] Remote Write: Fix signing for AWS sigv4 transport #13497
+* [BUGFIX] Observability: Exemplars emitted by Prometheus use "trace_id" not "traceID" #13589
+
+## 2.50.1 / 2024-02-26
+
+* [BUGFIX] API: Fix metadata API using wrong field names. #13633
+
+## 2.50.0 / 2024-02-22
+
+* [CHANGE] Remote Write: Error `storage.ErrTooOldSample` is now generating HTTP error 400 instead of HTTP error 500. #13335
+* [FEATURE] Remote Write: Drop old inmemory samples. Activated using the config entry `sample_age_limit`. #13002
+* [FEATURE] **Experimental**: Add support for ingesting zeros as created timestamps. (enabled under the feature-flag `created-timestamp-zero-ingestion`). #12733 #13279
+* [FEATURE] Promtool: Add `analyze` histograms command. #12331
+* [FEATURE] TSDB/compaction: Add a way to enable overlapping compaction. #13282 #13393 #13398
+* [FEATURE] Add automatic memory limit handling. Activated using the feature flag. `auto-gomemlimit` #13395
+* [ENHANCEMENT] Promtool: allow specifying multiple matchers in `promtool tsdb dump`. #13296
+* [ENHANCEMENT] PromQL: Restore more efficient version of `NewPossibleNonCounterInfo` annotation. #13022
+* [ENHANCEMENT] Kuma SD: Extend configuration to allow users to specify client ID. #13278
+* [ENHANCEMENT] PromQL: Use natural sort in `sort_by_label` and `sort_by_label_desc`. This is **experimental**. #13411
+* [ENHANCEMENT] Native Histograms: support `native_histogram_min_bucket_factor` in scrape_config. #13222
+* [ENHANCEMENT] Native Histograms: Issue warning if histogramRate is applied to the wrong kind of histogram. #13392
+* [ENHANCEMENT] TSDB: Make transaction isolation data structures smaller. #13015
+* [ENHANCEMENT] TSDB/postings: Optimize merge using Loser Tree. #12878
+* [ENHANCEMENT] TSDB: Simplify internal series delete function. #13261
+* [ENHANCEMENT] Agent: Performance improvement by making the global hash lookup table smaller. #13262
+* [ENHANCEMENT] PromQL: faster execution of metric functions, e.g. abs(), rate() #13446
+* [ENHANCEMENT] TSDB: Optimize label values with matchers by taking shortcuts. #13426
+* [ENHANCEMENT] Kubernetes SD: Check preconditions earlier and avoid unnecessary checks or iterations in kube_sd. #13408
+* [ENHANCEMENT] Promtool: Improve visibility for `promtool test rules` with JSON colored formatting. #13342
+* [ENHANCEMENT] Consoles: Exclude iowait and steal from CPU Utilisation. #9593
+* [ENHANCEMENT] Various improvements and optimizations on Native Histograms. #13267, #13215, #13276 #13289, #13340
+* [BUGFIX] Scraping: Fix quality value in HTTP Accept header. #13313
+* [BUGFIX] UI: Fix usage of the function `time()` that was crashing. #13371
+* [BUGFIX] Azure SD: Fix SD crashing when it finds a VM scale set. #13578
+
+## 2.49.1 / 2024-01-15
+
+* [BUGFIX] TSDB: Fixed a wrong `q=` value in scrape accept header #13313
+
+## 2.49.0 / 2024-01-15
+
+* [FEATURE] Promtool: Add `--run` flag promtool test rules command. #12206
+* [FEATURE] SD: Add support for `NS` records to DNS SD. #13219
+* [FEATURE] UI: Add heatmap visualization setting in the Graph tab, useful histograms. #13096 #13371
+* [FEATURE] Scraping: Add `scrape_config.enable_compression` (default true) to disable gzip compression when scraping the target. #13166
+* [FEATURE] PromQL: Add a `promql-experimental-functions` feature flag containing some new experimental PromQL functions. #13103 NOTE: More experimental functions might be added behind the same feature flag in the future. Added functions:
+  * Experimental `mad_over_time` (median absolute deviation around the median) function. #13059
+  * Experimental `sort_by_label` and `sort_by_label_desc` functions allowing sorting returned series by labels. #11299
+* [FEATURE] SD: Add `__meta_linode_gpus` label to Linode SD. #13097
+* [FEATURE] API: Add `exclude_alerts` query parameter to `/api/v1/rules` to only return recording rules. #12999
+* [FEATURE] TSDB: --storage.tsdb.retention.time flag value is now exposed as a `prometheus_tsdb_retention_limit_seconds` metric. #12986
+* [FEATURE] Scraping: Add ability to specify priority of scrape protocols to accept during scrape (e.g. to scrape Prometheus proto format for certain jobs). This can be changed by setting `global.scrape_protocols` and `scrape_config.scrape_protocols`. #12738
+* [ENHANCEMENT] Scraping: Automated handling of scraping histograms that violate `scrape_config.native_histogram_bucket_limit` setting. #13129
+* [ENHANCEMENT] Scraping: Optimized memory allocations when scraping. #12992
+* [ENHANCEMENT] SD: Added cache for Azure SD to avoid rate-limits. #12622
+* [ENHANCEMENT] TSDB: Various improvements to OOO exemplar scraping. E.g. allowing ingestion of exemplars with the same timestamp, but with different labels. #13021
+* [ENHANCEMENT] API: Optimize `/api/v1/labels` and `/api/v1/label/<label_name>/values` when 1 set of matchers are used. #12888
+* [ENHANCEMENT] TSDB: Various optimizations for TSDB block index, head mmap chunks and WAL, reducing latency and memory allocations (improving API calls, compaction queries etc). #12997 #13058 #13056 #13040
+* [ENHANCEMENT] PromQL: Optimize memory allocations and latency when querying float histograms. #12954
+* [ENHANCEMENT] Rules: Instrument TraceID in log lines for rule evaluations. #13034
+* [ENHANCEMENT] PromQL: Optimize memory allocations in query_range calls. #13043
+* [ENHANCEMENT] Promtool: unittest interval now defaults to evaluation_intervals when not set. #12729
+* [BUGFIX] SD: Fixed Azure SD public IP reporting #13241
+* [BUGFIX] API: Fix inaccuracies in posting cardinality statistics. #12653
+* [BUGFIX] PromQL: Fix inaccuracies of `histogram_quantile` with classic histograms. #13153
+* [BUGFIX] TSDB: Fix rare fails or inaccurate queries with OOO samples. #13115
+* [BUGFIX] TSDB: Fix rare panics on append commit when exemplars are used. #13092
+* [BUGFIX] TSDB: Fix exemplar WAL storage, so remote write can send/receive samples before exemplars. #13113
+* [BUGFIX] Mixins: Fix `url` filter on remote write dashboards. #10721
+* [BUGFIX] PromQL/TSDB: Various fixes to float histogram operations. #12891 #12977 #12609 #13190 #13189 #13191 #13201 #13212 #13208
+* [BUGFIX] Promtool: Fix int32 overflow issues for 32-bit architectures. #12978
+* [BUGFIX] SD: Fix Azure VM Scale Set NIC issue. #13283
+
+## 2.48.1 / 2023-12-07
+
+* [BUGFIX] TSDB: Make the wlog watcher read segments synchronously when not tailing. #13224
+* [BUGFIX] Agent: Participate in notify calls (fixes slow down in remote write handling introduced in 2.45). #13223
 
 ## 2.48.0 / 2023-11-16
 

CONTRIBUTING.md

@@ -95,7 +95,7 @@ can modify the `./promql/parser/generated_parser.y.go` manually.
 ```golang
 // As of writing this was somewhere around line 600.
 var (
-	yyDebug        = 0 // This can be be a number 0 -> 5.
+	yyDebug        = 0 // This can be a number 0 -> 5.
 	yyErrorVerbose = false  // This can be set to true.
 )
 

MAINTAINERS.md

@@ -1,15 +1,23 @@
 # Maintainers
 
-Julien Pivotto (<roidelapluie@prometheus.io> / @roidelapluie) and Levi Harrison (<levi@leviharrison.dev> / @LeviHarrison) are the main/default maintainers, some parts of the codebase have other maintainers:
+General maintainers:
+* Bryan Boreham (bjboreham@gmail.com / @bboreham)
+* Levi Harrison (levi@leviharrison.dev / @LeviHarrison)
+* Ayoub Mrini (ayoubmrini424@gmail.com / @machine424)
+* Julien Pivotto (roidelapluie@prometheus.io / @roidelapluie)
 
+Maintainers for specific parts of the codebase:
 * `cmd`
   * `promtool`: David Leadbeater (<dgl@dgl.cx> / @dgl)
 * `discovery`
   * `k8s`: Frederic Branczyk (<fbranczyk@gmail.com> / @brancz)
 * `documentation`
   * `prometheus-mixin`: Matthias Loibl (<mail@matthiasloibl.com> / @metalmatze)
+* `model/histogram` and other code related to native histograms: Björn Rabenstein (<beorn@grafana.com> / @beorn7), 
+George Krajcsovits (<gyorgy.krajcsovits@grafana.com> / @krajorama)
 * `storage`
-  * `remote`: Chris Marchbanks (<csmarchbanks@gmail.com> / @csmarchbanks), Callum Styan (<callumstyan@gmail.com> / @cstyan), Bartłomiej Płotka (<bwplotka@gmail.com> / @bwplotka), Tom Wilkie (<tom.wilkie@gmail.com> / @tomwilkie)
+  * `remote`: Callum Styan (<callumstyan@gmail.com> / @cstyan), Bartłomiej Płotka (<bwplotka@gmail.com> / @bwplotka), Tom Wilkie (<tom.wilkie@gmail.com> / @tomwilkie)
+    * `otlptranslator`: Arve Knudsen (<arve.knudsen@gmail.com> / @aknuds1), Jesús Vázquez (<jesus.vazquez@grafana.com> / @jesusvazquez)
 * `tsdb`: Ganesh Vernekar (<ganesh@grafana.com> / @codesome), Bartłomiej Płotka (<bwplotka@gmail.com> / @bwplotka), Jesús Vázquez (<jesus.vazquez@grafana.com> / @jesusvazquez)
   * `agent`: Robert Fratto (<robert.fratto@grafana.com> / @rfratto)
 * `web`
@@ -22,3 +30,13 @@ size of this repository, the natural changes in focus of maintainers over time,
 and nuances of where particular features live, this list will always be
 incomplete and out of date. However the listed maintainer(s) should be able to
 direct a PR/question to the right person.
+
+v3 release coordinators:
+* Alex Greenbank (<alex.greenbank@grafana.com> / @alexgreenbank)
+* Carrie Edwards (<carrie.edwards@grafana.com> / @carrieedwards)
+* Fiona Liao (<fiona.liao@grafana.com> / @fionaliao)
+* Jan Fajerski (<github@fajerski.name> / @jan--f)
+* Jesús Vázquez (<jesus.vazquez@grafana.com> / @jesusvazquez)
+* Nico Pazos (<nicolas.pazos-mendez@grafana.com> / @npazosmendez)
+* Owen Williams (<owen.williams@grafana.com> / @ywwg)
+* Tom Braack (<me@shorez.de> / @sh0rez)

Makefile

@@ -24,6 +24,7 @@ TSDB_BENCHMARK_DATASET ?= ./tsdb/testdata/20kseries.json
 TSDB_BENCHMARK_OUTPUT_DIR ?= ./benchout
 
 GOLANGCI_LINT_OPTS ?= --timeout 4m
+GOYACC_VERSION ?= v0.6.0
 
 include Makefile.common
 
@@ -78,24 +79,42 @@ assets-tarball: assets
 	@echo '>> packaging assets'
 	scripts/package_assets.sh
 
-# We only want to generate the parser when there's changes to the grammar.
 .PHONY: parser
 parser:
 	@echo ">> running goyacc to generate the .go file."
-ifeq (, $(shell command -v goyacc > /dev/null))
+ifeq (, $(shell command -v goyacc 2> /dev/null))
 	@echo "goyacc not installed so skipping"
-	@echo "To install: go install golang.org/x/tools/cmd/goyacc@v0.6.0"
+	@echo "To install: \"go install golang.org/x/tools/cmd/goyacc@$(GOYACC_VERSION)\" or run \"make install-goyacc\""
 else
-	goyacc -o promql/parser/generated_parser.y.go promql/parser/generated_parser.y
+	$(MAKE) promql/parser/generated_parser.y.go
 endif
 
+promql/parser/generated_parser.y.go: promql/parser/generated_parser.y
+	@echo ">> running goyacc to generate the .go file."
+	@goyacc -l -o promql/parser/generated_parser.y.go promql/parser/generated_parser.y
+
+.PHONY: clean-parser
+clean-parser:
+	@echo ">> cleaning generated parser"
+	@rm -f promql/parser/generated_parser.y.go
+
+.PHONY: check-generated-parser
+check-generated-parser: clean-parser promql/parser/generated_parser.y.go
+	@echo ">> checking generated parser"
+	@git diff --exit-code -- promql/parser/generated_parser.y.go || (echo "Generated parser is out of date. Please run 'make parser' and commit the changes." && false)
+
+.PHONY: install-goyacc
+install-goyacc:
+	@echo ">> installing goyacc $(GOYACC_VERSION)"
+	@go install golang.org/x/tools/cmd/goyacc@$(GOYACC_VERSION)
+
 .PHONY: test
 # If we only want to only test go code we have to change the test target
 # which is called by all.
 ifeq ($(GO_ONLY),1)
 test: common-test check-go-mod-version
 else
-test: common-test ui-build-module ui-test ui-lint check-go-mod-version
+test: check-generated-parser common-test ui-build-module ui-test ui-lint check-go-mod-version
 endif
 
 .PHONY: npm_licenses

Makefile.common

@@ -49,20 +49,20 @@ endif
 GOTEST := $(GO) test
 GOTEST_DIR :=
 ifneq ($(CIRCLE_JOB),)
-ifneq ($(shell command -v gotestsum > /dev/null),)
+ifneq ($(shell command -v gotestsum 2> /dev/null),)
 	GOTEST_DIR := test-results
 	GOTEST := gotestsum --junitfile $(GOTEST_DIR)/unit-tests.xml --
 endif
 endif
 
-PROMU_VERSION ?= 0.15.0
+PROMU_VERSION ?= 0.17.0
 PROMU_URL     := https://github.com/prometheus/promu/releases/download/v$(PROMU_VERSION)/promu-$(PROMU_VERSION).$(GO_BUILD_PLATFORM).tar.gz
 
 SKIP_GOLANGCI_LINT :=
 GOLANGCI_LINT :=
 GOLANGCI_LINT_OPTS ?=
-GOLANGCI_LINT_VERSION ?= v1.55.2
-# golangci-lint only supports linux, darwin and windows platforms on i386/amd64 and arm64.
+GOLANGCI_LINT_VERSION ?= v1.56.2
+# golangci-lint only supports linux, darwin and windows platforms on i386/amd64/arm64.
 # windows isn't included here because of the path separator being different.
 ifeq ($(GOHOSTOS),$(filter $(GOHOSTOS),linux darwin))
 	ifeq ($(GOHOSTARCH),$(filter $(GOHOSTARCH),amd64 i386 arm64))
@@ -169,16 +169,20 @@ common-vet:
 common-lint: $(GOLANGCI_LINT)
 ifdef GOLANGCI_LINT
 	@echo ">> running golangci-lint"
-# 'go list' needs to be executed before staticcheck to prepopulate the modules cache.
-# Otherwise staticcheck might fail randomly for some reason not yet explained.
-	$(GO) list -e -compiled -test=true -export=false -deps=true -find=false -tags= -- ./... > /dev/null
 	$(GOLANGCI_LINT) run $(GOLANGCI_LINT_OPTS) $(pkgs)
 endif
 
+.PHONY: common-lint-fix
+common-lint-fix: $(GOLANGCI_LINT)
+ifdef GOLANGCI_LINT
+	@echo ">> running golangci-lint fix"
+	$(GOLANGCI_LINT) run --fix $(GOLANGCI_LINT_OPTS) $(pkgs)
+endif
+
 .PHONY: common-yamllint
 common-yamllint:
 	@echo ">> running yamllint on all YAML files in the repository"
-ifeq (, $(shell command -v yamllint > /dev/null))
+ifeq (, $(shell command -v yamllint 2> /dev/null))
 	@echo "yamllint not installed so skipping"
 else
 	yamllint .
@@ -204,6 +208,10 @@ common-tarball: promu
 	@echo ">> building release tarball"
 	$(PROMU) tarball --prefix $(PREFIX) $(BIN_DIR)
 
+.PHONY: common-docker-repo-name
+common-docker-repo-name:
+	@echo "$(DOCKER_REPO)/$(DOCKER_IMAGE_NAME)"
+
 .PHONY: common-docker $(BUILD_DOCKER_ARCHS)
 common-docker: $(BUILD_DOCKER_ARCHS)
 $(BUILD_DOCKER_ARCHS): common-docker-%:

README.md

@@ -14,7 +14,7 @@ examples and guides.</p>
 [![CII Best Practices](https://bestpractices.coreinfrastructure.org/projects/486/badge)](https://bestpractices.coreinfrastructure.org/projects/486)
 [![Gitpod ready-to-code](https://img.shields.io/badge/Gitpod-ready--to--code-blue?logo=gitpod)](https://gitpod.io/#https://github.com/prometheus/prometheus)
 [![Fuzzing Status](https://oss-fuzz-build-logs.storage.googleapis.com/badges/prometheus.svg)](https://bugs.chromium.org/p/oss-fuzz/issues/list?sort=-opened&can=1&q=proj:prometheus)
-[![OpenSSF Scorecard](https://api.securityscorecards.dev/projects/github.com/prometheus/prometheus/badge)](https://api.securityscorecards.dev/projects/github.com/prometheus/prometheus)
+[![OpenSSF Scorecard](https://api.securityscorecards.dev/projects/github.com/prometheus/prometheus/badge)](https://securityscorecards.dev/viewer/?uri=github.com/prometheus/prometheus)
 
 </div>
 
@@ -149,7 +149,7 @@ We are publishing our Remote Write protobuf independently at
 You can use that as a library:
 
 ```shell
-go get go.buf.build/protocolbuffers/go/prometheus/prometheus
+go get buf.build/gen/go/prometheus/prometheus/protocolbuffers/go@latest
 ```
 
 This is experimental.

RELEASE.md

@@ -54,7 +54,10 @@ Release cadence of first pre-releases being cut is 6 weeks.
 | v2.47          | 2023-08-23                                 | Bryan Boreham (GitHub: @bboreham)           |
 | v2.48          | 2023-10-04                                 | Levi Harrison (GitHub: @LeviHarrison)       |
 | v2.49          | 2023-12-05                                 | Bartek Plotka (GitHub: @bwplotka)           |
-| v2.50          | 2024-01-16                                 | **searching for volunteer**                 |
+| v2.50          | 2024-01-16                                 | Augustin Husson (GitHub: @nexucis)          |
+| v2.51          | 2024-03-07                                 | Bryan Boreham (GitHub: @bboreham)           |
+| v2.52          | 2024-04-22                                 | Arthur Silva Sens (GitHub: @ArthurSens)     |
+| v2.53          | 2024-06-03                                 | George Krajcsovits (GitHub: @krajorama)     |
 
 If you are interested in volunteering please create a pull request against the [prometheus/prometheus](https://github.com/prometheus/prometheus) repository and propose yourself for the release series of your choice.
 

VERSION

@@ -1 +1 @@
-2.48.0
+2.52.0-rc.1

cmd/prometheus/main.go

@@ -33,6 +33,7 @@ import (
 	"syscall"
 	"time"
 
+	"github.com/KimMachineGun/automemlimit/memlimit"
 	"github.com/alecthomas/kingpin/v2"
 	"github.com/alecthomas/units"
 	"github.com/go-kit/log"
@@ -41,6 +42,7 @@ import (
 	"github.com/mwitkow/go-conntrack"
 	"github.com/oklog/run"
 	"github.com/prometheus/client_golang/prometheus"
+	versioncollector "github.com/prometheus/client_golang/prometheus/collectors/version"
 	"github.com/prometheus/common/model"
 	"github.com/prometheus/common/promlog"
 	promlogflag "github.com/prometheus/common/promlog/flag"
@@ -98,7 +100,7 @@ var (
 )
 
 func init() {
-	prometheus.MustRegister(version.NewCollector(strings.ReplaceAll(appName, "-", "_")))
+	prometheus.MustRegister(versioncollector.NewCollector(strings.ReplaceAll(appName, "-", "_")))
 
 	var err error
 	defaultRetentionDuration, err = model.ParseDuration(defaultRetentionString)
@@ -136,6 +138,7 @@ type flagConfig struct {
 	forGracePeriod      model.Duration
 	outageTolerance     model.Duration
 	resendDelay         model.Duration
+	maxConcurrentEvals  int64
 	web                 web.Options
 	scrape              scrape.Options
 	tsdb                tsdbOptions
@@ -147,13 +150,16 @@ type flagConfig struct {
 	queryMaxSamples     int
 	RemoteFlushDeadline model.Duration
 
-	featureList []string
+	featureList   []string
+	memlimitRatio float64
 	// These options are extracted from featureList
 	// for ease of use.
 	enableExpandExternalLabels bool
 	enableNewSDManager         bool
 	enablePerStepStats         bool
 	enableAutoGOMAXPROCS       bool
+	enableAutoGOMEMLIMIT       bool
+	enableConcurrentRuleEval   bool
 
 	prometheusURL   string
 	corsRegexString string
@@ -197,6 +203,12 @@ func (c *flagConfig) setFeatureListOptions(logger log.Logger) error {
 			case "auto-gomaxprocs":
 				c.enableAutoGOMAXPROCS = true
 				level.Info(logger).Log("msg", "Automatically set GOMAXPROCS to match Linux container CPU quota")
+			case "auto-gomemlimit":
+				c.enableAutoGOMEMLIMIT = true
+				level.Info(logger).Log("msg", "Automatically set GOMEMLIMIT to match Linux container or system memory limit")
+			case "concurrent-rule-eval":
+				c.enableConcurrentRuleEval = true
+				level.Info(logger).Log("msg", "Experimental concurrent rule evaluation enabled.")
 			case "no-default-scrape-port":
 				c.scrape.NoDefaultPort = true
 				level.Info(logger).Log("msg", "No default port will be appended to scrape targets' addresses.")
@@ -205,6 +217,7 @@ func (c *flagConfig) setFeatureListOptions(logger log.Logger) error {
 				level.Info(logger).Log("msg", "Experimental PromQL functions enabled.")
 			case "native-histograms":
 				c.tsdb.EnableNativeHistograms = true
+				c.scrape.EnableNativeHistogramsIngestion = true
 				// Change relevant global variables. Hacky, but it's hard to pass a new option or default to unmarshallers.
 				config.DefaultConfig.GlobalConfig.ScrapeProtocols = config.DefaultProtoFirstScrapeProtocols
 				config.DefaultGlobalConfig.ScrapeProtocols = config.DefaultProtoFirstScrapeProtocols
@@ -262,6 +275,9 @@ func main() {
 	a.Flag("web.listen-address", "Address to listen on for UI, API, and telemetry.").
 		Default("0.0.0.0:9090").StringVar(&cfg.web.ListenAddress)
 
+	a.Flag("auto-gomemlimit.ratio", "The ratio of reserved GOMEMLIMIT memory to the detected maximum container or system memory").
+		Default("0.9").FloatVar(&cfg.memlimitRatio)
+
 	webConfig := a.Flag(
 		"web.config.file",
 		"[EXPERIMENTAL] Path to configuration file that can enable TLS or authentication.",
@@ -402,6 +418,9 @@ func main() {
 	serverOnlyFlag(a, "rules.alert.resend-delay", "Minimum amount of time to wait before resending an alert to Alertmanager.").
 		Default("1m").SetValue(&cfg.resendDelay)
 
+	serverOnlyFlag(a, "rules.max-concurrent-evals", "Global concurrency limit for independent rules that can run concurrently. When set, \"query.max-concurrency\" may need to be adjusted accordingly.").
+		Default("4").Int64Var(&cfg.maxConcurrentEvals)
+
 	a.Flag("scrape.adjust-timestamps", "Adjust scrape timestamps by up to `scrape.timestamp-tolerance` to align them to the intended schedule. See https://github.com/prometheus/prometheus/issues/7846 for more context. Experimental. This flag will be removed in a future release.").
 		Hidden().Default("true").BoolVar(&scrape.AlignScrapeTimestamps)
 
@@ -429,7 +448,7 @@ func main() {
 	a.Flag("scrape.discovery-reload-interval", "Interval used by scrape manager to throttle target groups updates.").
 		Hidden().Default("5s").SetValue(&cfg.scrape.DiscoveryReloadInterval)
 
-	a.Flag("enable-feature", "Comma separated feature names to enable. Valid options: agent, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-at-modifier, promql-negative-offset, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details.").
+	a.Flag("enable-feature", "Comma separated feature names to enable. Valid options: agent, auto-gomemlimit, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver, created-timestamp-zero-ingestion, concurrent-rule-eval. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details.").
 		Default("").StringsVar(&cfg.featureList)
 
 	promlogflag.AddFlags(a, &cfg.promlogConfig)
@@ -467,6 +486,11 @@ func main() {
 		os.Exit(3)
 	}
 
+	if cfg.memlimitRatio <= 0.0 || cfg.memlimitRatio > 1.0 {
+		fmt.Fprintf(os.Stderr, "--auto-gomemlimit.ratio must be greater than 0 and less than or equal to 1.")
+		os.Exit(1)
+	}
+
 	localStoragePath := cfg.serverStoragePath
 	if agentMode {
 		localStoragePath = cfg.agentStoragePath
@@ -630,9 +654,16 @@ func main() {
 		level.Error(logger).Log("msg", "failed to register Kubernetes client metrics", "err", err)
 		os.Exit(1)
 	}
+
+	sdMetrics, err := discovery.CreateAndRegisterSDMetrics(prometheus.DefaultRegisterer)
+	if err != nil {
+		level.Error(logger).Log("msg", "failed to register service discovery metrics", "err", err)
+		os.Exit(1)
+	}
+
 	if cfg.enableNewSDManager {
 		{
-			discMgr := discovery.NewManager(ctxScrape, log.With(logger, "component", "discovery manager scrape"), prometheus.DefaultRegisterer, discovery.Name("scrape"))
+			discMgr := discovery.NewManager(ctxScrape, log.With(logger, "component", "discovery manager scrape"), prometheus.DefaultRegisterer, sdMetrics, discovery.Name("scrape"))
 			if discMgr == nil {
 				level.Error(logger).Log("msg", "failed to create a discovery manager scrape")
 				os.Exit(1)
@@ -641,7 +672,7 @@ func main() {
 		}
 
 		{
-			discMgr := discovery.NewManager(ctxNotify, log.With(logger, "component", "discovery manager notify"), prometheus.DefaultRegisterer, discovery.Name("notify"))
+			discMgr := discovery.NewManager(ctxNotify, log.With(logger, "component", "discovery manager notify"), prometheus.DefaultRegisterer, sdMetrics, discovery.Name("notify"))
 			if discMgr == nil {
 				level.Error(logger).Log("msg", "failed to create a discovery manager notify")
 				os.Exit(1)
@@ -650,7 +681,7 @@ func main() {
 		}
 	} else {
 		{
-			discMgr := legacymanager.NewManager(ctxScrape, log.With(logger, "component", "discovery manager scrape"), prometheus.DefaultRegisterer, legacymanager.Name("scrape"))
+			discMgr := legacymanager.NewManager(ctxScrape, log.With(logger, "component", "discovery manager scrape"), prometheus.DefaultRegisterer, sdMetrics, legacymanager.Name("scrape"))
 			if discMgr == nil {
 				level.Error(logger).Log("msg", "failed to create a discovery manager scrape")
 				os.Exit(1)
@@ -659,7 +690,7 @@ func main() {
 		}
 
 		{
-			discMgr := legacymanager.NewManager(ctxNotify, log.With(logger, "component", "discovery manager notify"), prometheus.DefaultRegisterer, legacymanager.Name("notify"))
+			discMgr := legacymanager.NewManager(ctxNotify, log.With(logger, "component", "discovery manager notify"), prometheus.DefaultRegisterer, sdMetrics, legacymanager.Name("notify"))
 			if discMgr == nil {
 				level.Error(logger).Log("msg", "failed to create a discovery manager notify")
 				os.Exit(1)
@@ -695,6 +726,20 @@ func main() {
 		}
 	}
 
+	if cfg.enableAutoGOMEMLIMIT {
+		if _, err := memlimit.SetGoMemLimitWithOpts(
+			memlimit.WithRatio(cfg.memlimitRatio),
+			memlimit.WithProvider(
+				memlimit.ApplyFallback(
+					memlimit.FromCgroup,
+					memlimit.FromSystem,
+				),
+			),
+		); err != nil {
+			level.Warn(logger).Log("component", "automemlimit", "msg", "Failed to set GOMEMLIMIT automatically", "err", err)
+		}
+	}
+
 	if !agentMode {
 		opts := promql.EngineOpts{
 			Logger:                   log.With(logger, "component", "query engine"),
@@ -714,17 +759,19 @@ func main() {
 		queryEngine = promql.NewEngine(opts)
 
 		ruleManager = rules.NewManager(&rules.ManagerOptions{
-			Appendable:      fanoutStorage,
-			Queryable:       localStorage,
-			QueryFunc:       rules.EngineQueryFunc(queryEngine, fanoutStorage),
-			NotifyFunc:      rules.SendAlerts(notifierManager, cfg.web.ExternalURL.String()),
-			Context:         ctxRule,
-			ExternalURL:     cfg.web.ExternalURL,
-			Registerer:      prometheus.DefaultRegisterer,
-			Logger:          log.With(logger, "component", "rule manager"),
-			OutageTolerance: time.Duration(cfg.outageTolerance),
-			ForGracePeriod:  time.Duration(cfg.forGracePeriod),
-			ResendDelay:     time.Duration(cfg.resendDelay),
+			Appendable:             fanoutStorage,
+			Queryable:              localStorage,
+			QueryFunc:              rules.EngineQueryFunc(queryEngine, fanoutStorage),
+			NotifyFunc:             rules.SendAlerts(notifierManager, cfg.web.ExternalURL.String()),
+			Context:                ctxRule,
+			ExternalURL:            cfg.web.ExternalURL,
+			Registerer:             prometheus.DefaultRegisterer,
+			Logger:                 log.With(logger, "component", "rule manager"),
+			OutageTolerance:        time.Duration(cfg.outageTolerance),
+			ForGracePeriod:         time.Duration(cfg.forGracePeriod),
+			ResendDelay:            time.Duration(cfg.resendDelay),
+			MaxConcurrentEvals:     cfg.maxConcurrentEvals,
+			ConcurrentEvalsEnabled: cfg.enableConcurrentRuleEval,
 		})
 	}
 
@@ -914,8 +961,8 @@ func main() {
 			func() error {
 				// Don't forget to release the reloadReady channel so that waiting blocks can exit normally.
 				select {
-				case <-term:
-					level.Warn(logger).Log("msg", "Received SIGTERM, exiting gracefully...")
+				case sig := <-term:
+					level.Warn(logger).Log("msg", "Received an OS signal, exiting gracefully...", "signal", sig.String())
 					reloadReady.Close()
 				case <-webHandler.Quit():
 					level.Warn(logger).Log("msg", "Received termination request via web service, exiting gracefully...")
@@ -1646,6 +1693,7 @@ func (opts tsdbOptions) ToTSDBOptions() tsdb.Options {
 		EnableMemorySnapshotOnShutdown: opts.EnableMemorySnapshotOnShutdown,
 		EnableNativeHistograms:         opts.EnableNativeHistograms,
 		OutOfOrderTimeWindow:           opts.OutOfOrderTimeWindow,
+		EnableOverlappingCompaction:    true,
 	}
 }
 

cmd/prometheus/main_test.go

@@ -24,6 +24,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"runtime"
+	"strconv"
 	"strings"
 	"syscall"
 	"testing"
@@ -126,12 +127,9 @@ func TestFailedStartupExitCode(t *testing.T) {
 	require.Error(t, err)
 
 	var exitError *exec.ExitError
-	if errors.As(err, &exitError) {
-		status := exitError.Sys().(syscall.WaitStatus)
-		require.Equal(t, expectedExitStatus, status.ExitStatus())
-	} else {
-		t.Errorf("unable to retrieve the exit status for prometheus: %v", err)
-	}
+	require.ErrorAs(t, err, &exitError)
+	status := exitError.Sys().(syscall.WaitStatus)
+	require.Equal(t, expectedExitStatus, status.ExitStatus())
 }
 
 type senderFunc func(alerts ...*notifier.Alert)
@@ -192,11 +190,9 @@ func TestSendAlerts(t *testing.T) {
 
 	for i, tc := range testCases {
 		tc := tc
-		t.Run(fmt.Sprintf("%d", i), func(t *testing.T) {
+		t.Run(strconv.Itoa(i), func(t *testing.T) {
 			senderFunc := senderFunc(func(alerts ...*notifier.Alert) {
-				if len(tc.in) == 0 {
-					t.Fatalf("sender called with 0 alert")
-				}
+				require.NotEmpty(t, tc.in, "sender called with 0 alert")
 				require.Equal(t, tc.exp, alerts)
 			})
 			rules.SendAlerts(senderFunc, "http://localhost:9090")(context.TODO(), "up", tc.in...)
@@ -228,7 +224,7 @@ func TestWALSegmentSizeBounds(t *testing.T) {
 			go func() { done <- prom.Wait() }()
 			select {
 			case err := <-done:
-				t.Errorf("prometheus should be still running: %v", err)
+				require.Fail(t, "prometheus should be still running: %v", err)
 			case <-time.After(startupTime):
 				prom.Process.Kill()
 				<-done
@@ -239,12 +235,9 @@ func TestWALSegmentSizeBounds(t *testing.T) {
 		err = prom.Wait()
 		require.Error(t, err)
 		var exitError *exec.ExitError
-		if errors.As(err, &exitError) {
-			status := exitError.Sys().(syscall.WaitStatus)
-			require.Equal(t, expectedExitStatus, status.ExitStatus())
-		} else {
-			t.Errorf("unable to retrieve the exit status for prometheus: %v", err)
-		}
+		require.ErrorAs(t, err, &exitError)
+		status := exitError.Sys().(syscall.WaitStatus)
+		require.Equal(t, expectedExitStatus, status.ExitStatus())
 	}
 }
 
@@ -274,7 +267,7 @@ func TestMaxBlockChunkSegmentSizeBounds(t *testing.T) {
 			go func() { done <- prom.Wait() }()
 			select {
 			case err := <-done:
-				t.Errorf("prometheus should be still running: %v", err)
+				require.Fail(t, "prometheus should be still running: %v", err)
 			case <-time.After(startupTime):
 				prom.Process.Kill()
 				<-done
@@ -285,12 +278,9 @@ func TestMaxBlockChunkSegmentSizeBounds(t *testing.T) {
 		err = prom.Wait()
 		require.Error(t, err)
 		var exitError *exec.ExitError
-		if errors.As(err, &exitError) {
-			status := exitError.Sys().(syscall.WaitStatus)
-			require.Equal(t, expectedExitStatus, status.ExitStatus())
-		} else {
-			t.Errorf("unable to retrieve the exit status for prometheus: %v", err)
-		}
+		require.ErrorAs(t, err, &exitError)
+		status := exitError.Sys().(syscall.WaitStatus)
+		require.Equal(t, expectedExitStatus, status.ExitStatus())
 	}
 }
 
@@ -347,10 +337,8 @@ func getCurrentGaugeValuesFor(t *testing.T, reg prometheus.Gatherer, metricNames
 			}
 
 			require.Len(t, g.GetMetric(), 1)
-			if _, ok := res[m]; ok {
-				t.Error("expected only one metric family for", m)
-				t.FailNow()
-			}
+			_, ok := res[m]
+			require.False(t, ok, "expected only one metric family for", m)
 			res[m] = *g.GetMetric()[0].GetGauge().Value
 		}
 	}

cmd/prometheus/main_unix_test.go

@@ -12,7 +12,6 @@
 // limitations under the License.
 //
 //go:build !windows
-// +build !windows
 
 package main
 
@@ -24,6 +23,8 @@ import (
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/require"
+
 	"github.com/prometheus/prometheus/util/testutil"
 )
 
@@ -38,9 +39,7 @@ func TestStartupInterrupt(t *testing.T) {
 
 	prom := exec.Command(promPath, "-test.main", "--config.file="+promConfig, "--storage.tsdb.path="+t.TempDir(), "--web.listen-address=0.0.0.0"+port)
 	err := prom.Start()
-	if err != nil {
-		t.Fatalf("execution error: %v", err)
-	}
+	require.NoError(t, err)
 
 	done := make(chan error, 1)
 	go func() {
@@ -69,14 +68,11 @@ Loop:
 		time.Sleep(500 * time.Millisecond)
 	}
 
-	if !startedOk {
-		t.Fatal("prometheus didn't start in the specified timeout")
-	}
-	switch err := prom.Process.Kill(); {
-	case err == nil:
-		t.Errorf("prometheus didn't shutdown gracefully after sending the Interrupt signal")
-	case stoppedErr != nil && stoppedErr.Error() != "signal: interrupt":
-		// TODO: find a better way to detect when the process didn't exit as expected!
-		t.Errorf("prometheus exited with an unexpected error: %v", stoppedErr)
+	require.True(t, startedOk, "prometheus didn't start in the specified timeout")
+	err = prom.Process.Kill()
+	require.Error(t, err, "prometheus didn't shutdown gracefully after sending the Interrupt signal")
+	// TODO - find a better way to detect when the process didn't exit as expected!
+	if stoppedErr != nil {
+		require.EqualError(t, stoppedErr, "signal: interrupt", "prometheus exit")
 	}
 }

cmd/promtool/analyze.go

@@ -0,0 +1,370 @@
+// Copyright 2023 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"math"
+	"net/http"
+	"net/url"
+	"os"
+	"sort"
+	"strconv"
+	"strings"
+	"time"
+
+	v1 "github.com/prometheus/client_golang/api/prometheus/v1"
+	"github.com/prometheus/common/model"
+
+	"github.com/prometheus/prometheus/model/labels"
+)
+
+var (
+	errNotNativeHistogram = fmt.Errorf("not a native histogram")
+	errNotEnoughData      = fmt.Errorf("not enough data")
+
+	outputHeader = `Bucket stats for each histogram series over time
+------------------------------------------------
+First the min, avg, and max number of populated buckets, followed by the total
+number of buckets (only if different from the max number of populated buckets
+which is typical for classic but not native histograms).`
+	outputFooter = `Aggregated bucket stats
+-----------------------
+Each line shows min/avg/max over the series above.`
+)
+
+type QueryAnalyzeConfig struct {
+	metricType string
+	duration   time.Duration
+	time       string
+	matchers   []string
+}
+
+// run retrieves metrics that look like conventional histograms (i.e. have _bucket
+// suffixes) or native histograms, depending on metricType flag.
+func (c *QueryAnalyzeConfig) run(url *url.URL, roundtripper http.RoundTripper) error {
+	if c.metricType != "histogram" {
+		return fmt.Errorf("analyze type is %s, must be 'histogram'", c.metricType)
+	}
+
+	ctx := context.Background()
+
+	api, err := newAPI(url, roundtripper, nil)
+	if err != nil {
+		return err
+	}
+
+	var endTime time.Time
+	if c.time != "" {
+		endTime, err = parseTime(c.time)
+		if err != nil {
+			return fmt.Errorf("error parsing time '%s': %w", c.time, err)
+		}
+	} else {
+		endTime = time.Now()
+	}
+
+	return c.getStatsFromMetrics(ctx, api, endTime, os.Stdout, c.matchers)
+}
+
+func (c *QueryAnalyzeConfig) getStatsFromMetrics(ctx context.Context, api v1.API, endTime time.Time, out io.Writer, matchers []string) error {
+	fmt.Fprintf(out, "%s\n\n", outputHeader)
+	metastatsNative := newMetaStatistics()
+	metastatsClassic := newMetaStatistics()
+	for _, matcher := range matchers {
+		seriesSel := seriesSelector(matcher, c.duration)
+		matrix, err := querySamples(ctx, api, seriesSel, endTime)
+		if err != nil {
+			return err
+		}
+
+		matrices := make(map[string]model.Matrix)
+		for _, series := range matrix {
+			// We do not handle mixed types. If there are float values, we assume it is a
+			// classic histogram, otherwise we assume it is a native histogram, and we
+			// ignore series with errors if they do not match the expected type.
+			if len(series.Values) == 0 {
+				stats, err := calcNativeBucketStatistics(series)
+				if err != nil {
+					if errors.Is(err, errNotNativeHistogram) || errors.Is(err, errNotEnoughData) {
+						continue
+					}
+					return err
+				}
+				fmt.Fprintf(out, "- %s (native): %v\n", series.Metric, *stats)
+				metastatsNative.update(stats)
+			} else {
+				lbs := model.LabelSet(series.Metric).Clone()
+				if _, ok := lbs["le"]; !ok {
+					continue
+				}
+				metricName := string(lbs[labels.MetricName])
+				if !strings.HasSuffix(metricName, "_bucket") {
+					continue
+				}
+				delete(lbs, labels.MetricName)
+				delete(lbs, "le")
+				key := formatSeriesName(metricName, lbs)
+				matrices[key] = append(matrices[key], series)
+			}
+		}
+
+		for key, matrix := range matrices {
+			stats, err := calcClassicBucketStatistics(matrix)
+			if err != nil {
+				if errors.Is(err, errNotEnoughData) {
+					continue
+				}
+				return err
+			}
+			fmt.Fprintf(out, "- %s (classic): %v\n", key, *stats)
+			metastatsClassic.update(stats)
+		}
+	}
+	fmt.Fprintf(out, "\n%s\n", outputFooter)
+	if metastatsNative.Count() > 0 {
+		fmt.Fprintf(out, "\nNative %s\n", metastatsNative)
+	}
+	if metastatsClassic.Count() > 0 {
+		fmt.Fprintf(out, "\nClassic %s\n", metastatsClassic)
+	}
+	return nil
+}
+
+func seriesSelector(metricName string, duration time.Duration) string {
+	builder := strings.Builder{}
+	builder.WriteString(metricName)
+	builder.WriteRune('[')
+	builder.WriteString(duration.String())
+	builder.WriteRune(']')
+	return builder.String()
+}
+
+func formatSeriesName(metricName string, lbs model.LabelSet) string {
+	builder := strings.Builder{}
+	builder.WriteString(metricName)
+	builder.WriteString(lbs.String())
+	return builder.String()
+}
+
+func querySamples(ctx context.Context, api v1.API, query string, end time.Time) (model.Matrix, error) {
+	values, _, err := api.Query(ctx, query, end)
+	if err != nil {
+		return nil, err
+	}
+
+	matrix, ok := values.(model.Matrix)
+	if !ok {
+		return nil, fmt.Errorf("query of buckets resulted in non-Matrix")
+	}
+
+	return matrix, nil
+}
+
+// minPop/avgPop/maxPop is for the number of populated (non-zero) buckets.
+// total is the total number of buckets across all samples in the series,
+// populated or not.
+type statistics struct {
+	minPop, maxPop, total int
+	avgPop                float64
+}
+
+func (s statistics) String() string {
+	if s.maxPop == s.total {
+		return fmt.Sprintf("%d/%.3f/%d", s.minPop, s.avgPop, s.maxPop)
+	}
+	return fmt.Sprintf("%d/%.3f/%d/%d", s.minPop, s.avgPop, s.maxPop, s.total)
+}
+
+func calcClassicBucketStatistics(matrix model.Matrix) (*statistics, error) {
+	numBuckets := len(matrix)
+
+	stats := &statistics{
+		minPop: math.MaxInt,
+		total:  numBuckets,
+	}
+
+	if numBuckets == 0 || len(matrix[0].Values) < 2 {
+		return stats, errNotEnoughData
+	}
+
+	numSamples := len(matrix[0].Values)
+
+	sortMatrix(matrix)
+
+	totalPop := 0
+	for timeIdx := 0; timeIdx < numSamples; timeIdx++ {
+		curr, err := getBucketCountsAtTime(matrix, numBuckets, timeIdx)
+		if err != nil {
+			return stats, err
+		}
+		countPop := 0
+		for _, b := range curr {
+			if b != 0 {
+				countPop++
+			}
+		}
+
+		totalPop += countPop
+		if stats.minPop > countPop {
+			stats.minPop = countPop
+		}
+		if stats.maxPop < countPop {
+			stats.maxPop = countPop
+		}
+	}
+	stats.avgPop = float64(totalPop) / float64(numSamples)
+	return stats, nil
+}
+
+func sortMatrix(matrix model.Matrix) {
+	sort.SliceStable(matrix, func(i, j int) bool {
+		return getLe(matrix[i]) < getLe(matrix[j])
+	})
+}
+
+func getLe(series *model.SampleStream) float64 {
+	lbs := model.LabelSet(series.Metric)
+	le, _ := strconv.ParseFloat(string(lbs["le"]), 64)
+	return le
+}
+
+func getBucketCountsAtTime(matrix model.Matrix, numBuckets, timeIdx int) ([]int, error) {
+	counts := make([]int, numBuckets)
+	if timeIdx >= len(matrix[0].Values) {
+		// Just return zeroes instead of erroring out so we can get partial results.
+		return counts, nil
+	}
+	counts[0] = int(matrix[0].Values[timeIdx].Value)
+	for i, bucket := range matrix[1:] {
+		if timeIdx >= len(bucket.Values) {
+			// Just return zeroes instead of erroring out so we can get partial results.
+			return counts, nil
+		}
+		curr := bucket.Values[timeIdx]
+		prev := matrix[i].Values[timeIdx]
+		// Assume the results are nicely aligned.
+		if curr.Timestamp != prev.Timestamp {
+			return counts, fmt.Errorf("matrix result is not time aligned")
+		}
+		counts[i+1] = int(curr.Value - prev.Value)
+	}
+	return counts, nil
+}
+
+type bucketBounds struct {
+	boundaries   int32
+	upper, lower float64
+}
+
+func makeBucketBounds(b *model.HistogramBucket) bucketBounds {
+	return bucketBounds{
+		boundaries: b.Boundaries,
+		upper:      float64(b.Upper),
+		lower:      float64(b.Lower),
+	}
+}
+
+func calcNativeBucketStatistics(series *model.SampleStream) (*statistics, error) {
+	stats := &statistics{
+		minPop: math.MaxInt,
+	}
+
+	overall := make(map[bucketBounds]struct{})
+	totalPop := 0
+	if len(series.Histograms) == 0 {
+		return nil, errNotNativeHistogram
+	}
+	if len(series.Histograms) == 1 {
+		return nil, errNotEnoughData
+	}
+	for _, histogram := range series.Histograms {
+		for _, bucket := range histogram.Histogram.Buckets {
+			bb := makeBucketBounds(bucket)
+			overall[bb] = struct{}{}
+		}
+		countPop := len(histogram.Histogram.Buckets)
+
+		totalPop += countPop
+		if stats.minPop > countPop {
+			stats.minPop = countPop
+		}
+		if stats.maxPop < countPop {
+			stats.maxPop = countPop
+		}
+	}
+	stats.avgPop = float64(totalPop) / float64(len(series.Histograms))
+	stats.total = len(overall)
+	return stats, nil
+}
+
+type distribution struct {
+	min, max, count int
+	avg             float64
+}
+
+func newDistribution() distribution {
+	return distribution{
+		min: math.MaxInt,
+	}
+}
+
+func (d *distribution) update(num int) {
+	if d.min > num {
+		d.min = num
+	}
+	if d.max < num {
+		d.max = num
+	}
+	d.count++
+	d.avg += float64(num)/float64(d.count) - d.avg/float64(d.count)
+}
+
+func (d distribution) String() string {
+	return fmt.Sprintf("%d/%.3f/%d", d.min, d.avg, d.max)
+}
+
+type metaStatistics struct {
+	minPop, avgPop, maxPop, total distribution
+}
+
+func newMetaStatistics() *metaStatistics {
+	return &metaStatistics{
+		minPop: newDistribution(),
+		avgPop: newDistribution(),
+		maxPop: newDistribution(),
+		total:  newDistribution(),
+	}
+}
+
+func (ms metaStatistics) Count() int {
+	return ms.minPop.count
+}
+
+func (ms metaStatistics) String() string {
+	if ms.maxPop == ms.total {
+		return fmt.Sprintf("histogram series (%d in total):\n- min populated: %v\n- avg populated: %v\n- max populated: %v", ms.Count(), ms.minPop, ms.avgPop, ms.maxPop)
+	}
+	return fmt.Sprintf("histogram series (%d in total):\n- min populated: %v\n- avg populated: %v\n- max populated: %v\n- total: %v", ms.Count(), ms.minPop, ms.avgPop, ms.maxPop, ms.total)
+}
+
+func (ms *metaStatistics) update(s *statistics) {
+	ms.minPop.update(s.minPop)
+	ms.avgPop.update(int(s.avgPop))
+	ms.maxPop.update(s.maxPop)
+	ms.total.update(s.total)
+}

cmd/promtool/analyze_test.go

@@ -0,0 +1,170 @@
+// Copyright 2023 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/prometheus/common/model"
+)
+
+var (
+	exampleMatrix = model.Matrix{
+		&model.SampleStream{
+			Metric: model.Metric{
+				"le": "+Inf",
+			},
+			Values: []model.SamplePair{
+				{
+					Value:     31,
+					Timestamp: 100,
+				},
+				{
+					Value:     32,
+					Timestamp: 200,
+				},
+				{
+					Value:     40,
+					Timestamp: 300,
+				},
+			},
+		},
+		&model.SampleStream{
+			Metric: model.Metric{
+				"le": "0.5",
+			},
+			Values: []model.SamplePair{
+				{
+					Value:     10,
+					Timestamp: 100,
+				},
+				{
+					Value:     11,
+					Timestamp: 200,
+				},
+				{
+					Value:     11,
+					Timestamp: 300,
+				},
+			},
+		},
+		&model.SampleStream{
+			Metric: model.Metric{
+				"le": "10",
+			},
+			Values: []model.SamplePair{
+				{
+					Value:     30,
+					Timestamp: 100,
+				},
+				{
+					Value:     31,
+					Timestamp: 200,
+				},
+				{
+					Value:     37,
+					Timestamp: 300,
+				},
+			},
+		},
+		&model.SampleStream{
+			Metric: model.Metric{
+				"le": "2",
+			},
+			Values: []model.SamplePair{
+				{
+					Value:     25,
+					Timestamp: 100,
+				},
+				{
+					Value:     26,
+					Timestamp: 200,
+				},
+				{
+					Value:     27,
+					Timestamp: 300,
+				},
+			},
+		},
+	}
+	exampleMatrixLength = len(exampleMatrix)
+)
+
+func init() {
+	sortMatrix(exampleMatrix)
+}
+
+func TestGetBucketCountsAtTime(t *testing.T) {
+	cases := []struct {
+		matrix   model.Matrix
+		length   int
+		timeIdx  int
+		expected []int
+	}{
+		{
+			exampleMatrix,
+			exampleMatrixLength,
+			0,
+			[]int{10, 15, 5, 1},
+		},
+		{
+			exampleMatrix,
+			exampleMatrixLength,
+			1,
+			[]int{11, 15, 5, 1},
+		},
+		{
+			exampleMatrix,
+			exampleMatrixLength,
+			2,
+			[]int{11, 16, 10, 3},
+		},
+	}
+
+	for _, c := range cases {
+		t.Run(fmt.Sprintf("exampleMatrix@%d", c.timeIdx), func(t *testing.T) {
+			res, err := getBucketCountsAtTime(c.matrix, c.length, c.timeIdx)
+			require.NoError(t, err)
+			require.Equal(t, c.expected, res)
+		})
+	}
+}
+
+func TestCalcClassicBucketStatistics(t *testing.T) {
+	cases := []struct {
+		matrix   model.Matrix
+		expected *statistics
+	}{
+		{
+			exampleMatrix,
+			&statistics{
+				minPop: 4,
+				avgPop: 4,
+				maxPop: 4,
+				total:  4,
+			},
+		},
+	}
+
+	for i, c := range cases {
+		t.Run(fmt.Sprintf("case %d", i), func(t *testing.T) {
+			res, err := calcClassicBucketStatistics(c.matrix)
+			require.NoError(t, err)
+			require.Equal(t, c.expected, res)
+		})
+	}
+}

cmd/promtool/archive.go

@@ -15,9 +15,10 @@ package main
 
 import (
 	"archive/tar"
-	"compress/gzip"
 	"fmt"
 	"os"
+
+	"github.com/klauspost/compress/gzip"
 )
 
 const filePerm = 0o666

cmd/promtool/backfill.go

@@ -88,7 +88,7 @@ func createBlocks(input []byte, mint, maxt, maxBlockDuration int64, maxSamplesIn
 	blockDuration := getCompatibleBlockDuration(maxBlockDuration)
 	mint = blockDuration * (mint / blockDuration)
 
-	db, err := tsdb.OpenDBReadOnly(outputDir, nil)
+	db, err := tsdb.OpenDBReadOnly(outputDir, "", nil)
 	if err != nil {
 		return err
 	}
@@ -127,7 +127,8 @@ func createBlocks(input []byte, mint, maxt, maxBlockDuration int64, maxSamplesIn
 
 			ctx := context.Background()
 			app := w.Appender(ctx)
-			p := textparse.NewOpenMetricsParser(input)
+			symbolTable := labels.NewSymbolTable() // One table per block means it won't grow too large.
+			p := textparse.NewOpenMetricsParser(input, symbolTable)
 			samplesCount := 0
 			for {
 				e, err := p.Next()
@@ -216,7 +217,7 @@ func createBlocks(input []byte, mint, maxt, maxBlockDuration int64, maxSamplesIn
 }
 
 func backfill(maxSamplesInAppender int, input []byte, outputDir string, humanReadable, quiet bool, maxBlockDuration time.Duration) (err error) {
-	p := textparse.NewOpenMetricsParser(input)
+	p := textparse.NewOpenMetricsParser(input, nil) // Don't need a SymbolTable to get max and min timestamps.
 	maxt, mint, err := getMinAndMaxTimestamps(p)
 	if err != nil {
 		return fmt.Errorf("getting min and max timestamp: %w", err)

cmd/promtool/backfill_test.go

@@ -26,6 +26,7 @@ import (
 	"github.com/prometheus/prometheus/storage"
 	"github.com/prometheus/prometheus/tsdb"
 	"github.com/prometheus/prometheus/tsdb/chunkenc"
+	"github.com/prometheus/prometheus/util/testutil"
 )
 
 type backfillSample struct {
@@ -76,7 +77,7 @@ func testBlocks(t *testing.T, db *tsdb.DB, expectedMinTime, expectedMaxTime, exp
 	allSamples := queryAllSeries(t, q, expectedMinTime, expectedMaxTime)
 	sortSamples(allSamples)
 	sortSamples(expectedSamples)
-	require.Equal(t, expectedSamples, allSamples, "did not create correct samples")
+	testutil.RequireEqual(t, expectedSamples, allSamples, "did not create correct samples")
 
 	if len(allSamples) > 0 {
 		require.Equal(t, expectedMinTime, allSamples[0].Timestamp, "timestamp of first sample is not the expected minimum time")

cmd/promtool/main.go

@@ -35,9 +35,7 @@ import (
 	"github.com/go-kit/log"
 	"github.com/google/pprof/profile"
 	"github.com/prometheus/client_golang/api"
-	v1 "github.com/prometheus/client_golang/api/prometheus/v1"
 	"github.com/prometheus/client_golang/prometheus"
-	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"github.com/prometheus/client_golang/prometheus/testutil/promlint"
 	config_util "github.com/prometheus/common/config"
 	"github.com/prometheus/common/model"
@@ -58,8 +56,8 @@ import (
 	"github.com/prometheus/prometheus/model/rulefmt"
 	"github.com/prometheus/prometheus/notifier"
 	_ "github.com/prometheus/prometheus/plugins" // Register plugins.
-	"github.com/prometheus/prometheus/promql"
 	"github.com/prometheus/prometheus/promql/parser"
+	"github.com/prometheus/prometheus/promql/promqltest"
 	"github.com/prometheus/prometheus/scrape"
 	"github.com/prometheus/prometheus/util/documentcli"
 )
@@ -185,6 +183,14 @@ func main() {
 	queryLabelsEnd := queryLabelsCmd.Flag("end", "End time (RFC3339 or Unix timestamp).").String()
 	queryLabelsMatch := queryLabelsCmd.Flag("match", "Series selector. Can be specified multiple times.").Strings()
 
+	queryAnalyzeCfg := &QueryAnalyzeConfig{}
+	queryAnalyzeCmd := queryCmd.Command("analyze", "Run queries against your Prometheus to analyze the usage pattern of certain metrics.")
+	queryAnalyzeCmd.Flag("server", "Prometheus server to query.").Required().URLVar(&serverURL)
+	queryAnalyzeCmd.Flag("type", "Type of metric: histogram.").Required().StringVar(&queryAnalyzeCfg.metricType)
+	queryAnalyzeCmd.Flag("duration", "Time frame to analyze.").Default("1h").DurationVar(&queryAnalyzeCfg.duration)
+	queryAnalyzeCmd.Flag("time", "Query time (RFC3339 or Unix timestamp), defaults to now.").StringVar(&queryAnalyzeCfg.time)
+	queryAnalyzeCmd.Flag("match", "Series selector. Can be specified multiple times.").Required().StringsVar(&queryAnalyzeCfg.matchers)
+
 	pushCmd := app.Command("push", "Push to a Prometheus server.")
 	pushCmd.Flag("http.config.file", "HTTP client configuration file for promtool to connect to Prometheus.").PlaceHolder("<filename>").ExistingFileVar(&httpConfigFilePath)
 	pushMetricsCmd := pushCmd.Command("metrics", "Push metrics to a prometheus remote write (for testing purpose only).")
@@ -204,6 +210,7 @@ func main() {
 		"test-rule-file",
 		"The unit test file.",
 	).Required().ExistingFiles()
+	testRulesDiff := testRulesCmd.Flag("diff", "[Experimental] Print colored differential output between expected & received output.").Default("false").Bool()
 
 	defaultDBPath := "data/"
 	tsdbCmd := app.Command("tsdb", "Run tsdb commands.")
@@ -228,9 +235,17 @@ func main() {
 
 	tsdbDumpCmd := tsdbCmd.Command("dump", "Dump samples from a TSDB.")
 	dumpPath := tsdbDumpCmd.Arg("db path", "Database path (default is "+defaultDBPath+").").Default(defaultDBPath).String()
+	dumpSandboxDirRoot := tsdbDumpCmd.Flag("sandbox-dir-root", "Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end.").Default(defaultDBPath).String()
 	dumpMinTime := tsdbDumpCmd.Flag("min-time", "Minimum timestamp to dump.").Default(strconv.FormatInt(math.MinInt64, 10)).Int64()
 	dumpMaxTime := tsdbDumpCmd.Flag("max-time", "Maximum timestamp to dump.").Default(strconv.FormatInt(math.MaxInt64, 10)).Int64()
-	dumpMatch := tsdbDumpCmd.Flag("match", "Series selector.").Default("{__name__=~'(?s:.*)'}").String()
+	dumpMatch := tsdbDumpCmd.Flag("match", "Series selector. Can be specified multiple times.").Default("{__name__=~'(?s:.*)'}").Strings()
+
+	tsdbDumpOpenMetricsCmd := tsdbCmd.Command("dump-openmetrics", "[Experimental] Dump samples from a TSDB into OpenMetrics text format, excluding native histograms and staleness markers, which are not representable in OpenMetrics.")
+	dumpOpenMetricsPath := tsdbDumpOpenMetricsCmd.Arg("db path", "Database path (default is "+defaultDBPath+").").Default(defaultDBPath).String()
+	dumpOpenMetricsSandboxDirRoot := tsdbDumpOpenMetricsCmd.Flag("sandbox-dir-root", "Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end.").Default(defaultDBPath).String()
+	dumpOpenMetricsMinTime := tsdbDumpOpenMetricsCmd.Flag("min-time", "Minimum timestamp to dump.").Default(strconv.FormatInt(math.MinInt64, 10)).Int64()
+	dumpOpenMetricsMaxTime := tsdbDumpOpenMetricsCmd.Flag("max-time", "Maximum timestamp to dump.").Default(strconv.FormatInt(math.MaxInt64, 10)).Int64()
+	dumpOpenMetricsMatch := tsdbDumpOpenMetricsCmd.Flag("match", "Series selector. Can be specified multiple times.").Default("{__name__=~'(?s:.*)'}").Strings()
 
 	importCmd := tsdbCmd.Command("create-blocks-from", "[Experimental] Import samples from input and produce TSDB blocks. Please refer to the storage docs for more details.")
 	importHumanReadable := importCmd.Flag("human-readable", "Print human readable values.").Short('r').Bool()
@@ -364,11 +379,12 @@ func main() {
 
 	case testRulesCmd.FullCommand():
 		os.Exit(RulesUnitTest(
-			promql.LazyLoaderOpts{
+			promqltest.LazyLoaderOpts{
 				EnableAtModifier:     true,
 				EnableNegativeOffset: true,
 			},
 			*testRulesRun,
+			*testRulesDiff,
 			*testRulesFiles...),
 		)
 
@@ -382,7 +398,9 @@ func main() {
 		os.Exit(checkErr(listBlocks(*listPath, *listHumanReadable)))
 
 	case tsdbDumpCmd.FullCommand():
-		os.Exit(checkErr(dumpSamples(ctx, *dumpPath, *dumpMinTime, *dumpMaxTime, *dumpMatch)))
+		os.Exit(checkErr(dumpSamples(ctx, *dumpPath, *dumpSandboxDirRoot, *dumpMinTime, *dumpMaxTime, *dumpMatch, formatSeriesSet)))
+	case tsdbDumpOpenMetricsCmd.FullCommand():
+		os.Exit(checkErr(dumpSamples(ctx, *dumpOpenMetricsPath, *dumpOpenMetricsSandboxDirRoot, *dumpOpenMetricsMinTime, *dumpOpenMetricsMaxTime, *dumpOpenMetricsMatch, formatSeriesSetOpenMetrics)))
 	// TODO(aSquare14): Work on adding support for custom block size.
 	case openMetricsImportCmd.FullCommand():
 		os.Exit(backfillOpenMetrics(*importFilePath, *importDBPath, *importHumanReadable, *importQuiet, *maxBlockDuration))
@@ -390,6 +408,9 @@ func main() {
 	case importRulesCmd.FullCommand():
 		os.Exit(checkErr(importRules(serverURL, httpRoundTripper, *importRulesStart, *importRulesEnd, *importRulesOutputDir, *importRulesEvalInterval, *maxBlockDuration, *importRulesFiles...)))
 
+	case queryAnalyzeCmd.FullCommand():
+		os.Exit(checkErr(queryAnalyzeCfg.run(serverURL, httpRoundTripper)))
+
 	case documentationCmd.FullCommand():
 		os.Exit(checkErr(documentcli.GenerateMarkdown(app.Model(), os.Stdout)))
 
@@ -463,7 +484,7 @@ func CheckServerStatus(serverURL *url.URL, checkEndpoint string, roundTripper ht
 		return err
 	}
 
-	request, err := http.NewRequest("GET", config.Address, nil)
+	request, err := http.NewRequest(http.MethodGet, config.Address, nil)
 	if err != nil {
 		return err
 	}
@@ -997,246 +1018,6 @@ func checkMetricsExtended(r io.Reader) ([]metricStat, int, error) {
 	return stats, total, nil
 }
 
-// QueryInstant performs an instant query against a Prometheus server.
-func QueryInstant(url *url.URL, roundTripper http.RoundTripper, query, evalTime string, p printer) int {
-	if url.Scheme == "" {
-		url.Scheme = "http"
-	}
-	config := api.Config{
-		Address:      url.String(),
-		RoundTripper: roundTripper,
-	}
-
-	// Create new client.
-	c, err := api.NewClient(config)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "error creating API client:", err)
-		return failureExitCode
-	}
-
-	eTime := time.Now()
-	if evalTime != "" {
-		eTime, err = parseTime(evalTime)
-		if err != nil {
-			fmt.Fprintln(os.Stderr, "error parsing evaluation time:", err)
-			return failureExitCode
-		}
-	}
-
-	// Run query against client.
-	api := v1.NewAPI(c)
-
-	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
-	val, _, err := api.Query(ctx, query, eTime) // Ignoring warnings for now.
-	cancel()
-	if err != nil {
-		return handleAPIError(err)
-	}
-
-	p.printValue(val)
-
-	return successExitCode
-}
-
-// QueryRange performs a range query against a Prometheus server.
-func QueryRange(url *url.URL, roundTripper http.RoundTripper, headers map[string]string, query, start, end string, step time.Duration, p printer) int {
-	if url.Scheme == "" {
-		url.Scheme = "http"
-	}
-	config := api.Config{
-		Address:      url.String(),
-		RoundTripper: roundTripper,
-	}
-
-	if len(headers) > 0 {
-		config.RoundTripper = promhttp.RoundTripperFunc(func(req *http.Request) (*http.Response, error) {
-			for key, value := range headers {
-				req.Header.Add(key, value)
-			}
-			return roundTripper.RoundTrip(req)
-		})
-	}
-
-	// Create new client.
-	c, err := api.NewClient(config)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "error creating API client:", err)
-		return failureExitCode
-	}
-
-	var stime, etime time.Time
-
-	if end == "" {
-		etime = time.Now()
-	} else {
-		etime, err = parseTime(end)
-		if err != nil {
-			fmt.Fprintln(os.Stderr, "error parsing end time:", err)
-			return failureExitCode
-		}
-	}
-
-	if start == "" {
-		stime = etime.Add(-5 * time.Minute)
-	} else {
-		stime, err = parseTime(start)
-		if err != nil {
-			fmt.Fprintln(os.Stderr, "error parsing start time:", err)
-			return failureExitCode
-		}
-	}
-
-	if !stime.Before(etime) {
-		fmt.Fprintln(os.Stderr, "start time is not before end time")
-		return failureExitCode
-	}
-
-	if step == 0 {
-		resolution := math.Max(math.Floor(etime.Sub(stime).Seconds()/250), 1)
-		// Convert seconds to nanoseconds such that time.Duration parses correctly.
-		step = time.Duration(resolution) * time.Second
-	}
-
-	// Run query against client.
-	api := v1.NewAPI(c)
-	r := v1.Range{Start: stime, End: etime, Step: step}
-	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
-	val, _, err := api.QueryRange(ctx, query, r) // Ignoring warnings for now.
-	cancel()
-
-	if err != nil {
-		return handleAPIError(err)
-	}
-
-	p.printValue(val)
-	return successExitCode
-}
-
-// QuerySeries queries for a series against a Prometheus server.
-func QuerySeries(url *url.URL, roundTripper http.RoundTripper, matchers []string, start, end string, p printer) int {
-	if url.Scheme == "" {
-		url.Scheme = "http"
-	}
-	config := api.Config{
-		Address:      url.String(),
-		RoundTripper: roundTripper,
-	}
-
-	// Create new client.
-	c, err := api.NewClient(config)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "error creating API client:", err)
-		return failureExitCode
-	}
-
-	stime, etime, err := parseStartTimeAndEndTime(start, end)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return failureExitCode
-	}
-
-	// Run query against client.
-	api := v1.NewAPI(c)
-	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
-	val, _, err := api.Series(ctx, matchers, stime, etime) // Ignoring warnings for now.
-	cancel()
-
-	if err != nil {
-		return handleAPIError(err)
-	}
-
-	p.printSeries(val)
-	return successExitCode
-}
-
-// QueryLabels queries for label values against a Prometheus server.
-func QueryLabels(url *url.URL, roundTripper http.RoundTripper, matchers []string, name, start, end string, p printer) int {
-	if url.Scheme == "" {
-		url.Scheme = "http"
-	}
-	config := api.Config{
-		Address:      url.String(),
-		RoundTripper: roundTripper,
-	}
-
-	// Create new client.
-	c, err := api.NewClient(config)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, "error creating API client:", err)
-		return failureExitCode
-	}
-
-	stime, etime, err := parseStartTimeAndEndTime(start, end)
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		return failureExitCode
-	}
-
-	// Run query against client.
-	api := v1.NewAPI(c)
-	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
-	val, warn, err := api.LabelValues(ctx, name, matchers, stime, etime)
-	cancel()
-
-	for _, v := range warn {
-		fmt.Fprintln(os.Stderr, "query warning:", v)
-	}
-	if err != nil {
-		return handleAPIError(err)
-	}
-
-	p.printLabelValues(val)
-	return successExitCode
-}
-
-func handleAPIError(err error) int {
-	var apiErr *v1.Error
-	if errors.As(err, &apiErr) && apiErr.Detail != "" {
-		fmt.Fprintf(os.Stderr, "query error: %v (detail: %s)\n", apiErr, strings.TrimSpace(apiErr.Detail))
-	} else {
-		fmt.Fprintln(os.Stderr, "query error:", err)
-	}
-
-	return failureExitCode
-}
-
-func parseStartTimeAndEndTime(start, end string) (time.Time, time.Time, error) {
-	var (
-		minTime = time.Now().Add(-9999 * time.Hour)
-		maxTime = time.Now().Add(9999 * time.Hour)
-		err     error
-	)
-
-	stime := minTime
-	etime := maxTime
-
-	if start != "" {
-		stime, err = parseTime(start)
-		if err != nil {
-			return stime, etime, fmt.Errorf("error parsing start time: %w", err)
-		}
-	}
-
-	if end != "" {
-		etime, err = parseTime(end)
-		if err != nil {
-			return stime, etime, fmt.Errorf("error parsing end time: %w", err)
-		}
-	}
-	return stime, etime, nil
-}
-
-func parseTime(s string) (time.Time, error) {
-	if t, err := strconv.ParseFloat(s, 64); err == nil {
-		s, ns := math.Modf(t)
-		return time.Unix(int64(s), int64(ns*float64(time.Second))).UTC(), nil
-	}
-	if t, err := time.Parse(time.RFC3339Nano, s); err == nil {
-		return t, nil
-	}
-	return time.Time{}, fmt.Errorf("cannot parse %q to a valid timestamp", s)
-}
-
 type endpointsGroup struct {
 	urlToFilename map[string]string
 	postProcess   func(b []byte) ([]byte, error)
@@ -1390,15 +1171,12 @@ func importRules(url *url.URL, roundTripper http.RoundTripper, start, end, outpu
 		evalInterval:     evalInterval,
 		maxBlockDuration: maxBlockDuration,
 	}
-	client, err := api.NewClient(api.Config{
-		Address:      url.String(),
-		RoundTripper: roundTripper,
-	})
+	api, err := newAPI(url, roundTripper, nil)
 	if err != nil {
 		return fmt.Errorf("new api client error: %w", err)
 	}
 
-	ruleImporter := newRuleImporter(log.NewLogfmtLogger(log.NewSyncWriter(os.Stderr)), cfg, v1.NewAPI(client))
+	ruleImporter := newRuleImporter(log.NewLogfmtLogger(log.NewSyncWriter(os.Stderr)), cfg, api)
 	errs := ruleImporter.loadGroups(ctx, files)
 	for _, err := range errs {
 		if err != nil {

cmd/promtool/main_test.go

@@ -25,6 +25,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"runtime"
+	"strconv"
 	"strings"
 	"syscall"
 	"testing"
@@ -410,7 +411,7 @@ func TestExitCodes(t *testing.T) {
 	} {
 		t.Run(c.file, func(t *testing.T) {
 			for _, lintFatal := range []bool{true, false} {
-				t.Run(fmt.Sprintf("%t", lintFatal), func(t *testing.T) {
+				t.Run(strconv.FormatBool(lintFatal), func(t *testing.T) {
 					args := []string{"-test.main", "check", "config", "testdata/" + c.file}
 					if lintFatal {
 						args = append(args, "--lint-fatal")

cmd/promtool/query.go

@@ -0,0 +1,251 @@
+// Copyright 2023 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package main
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"math"
+	"net/http"
+	"net/url"
+	"os"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/prometheus/client_golang/api"
+	v1 "github.com/prometheus/client_golang/api/prometheus/v1"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+
+	_ "github.com/prometheus/prometheus/plugins" // Register plugins.
+)
+
+func newAPI(url *url.URL, roundTripper http.RoundTripper, headers map[string]string) (v1.API, error) {
+	if url.Scheme == "" {
+		url.Scheme = "http"
+	}
+	config := api.Config{
+		Address:      url.String(),
+		RoundTripper: roundTripper,
+	}
+
+	if len(headers) > 0 {
+		config.RoundTripper = promhttp.RoundTripperFunc(func(req *http.Request) (*http.Response, error) {
+			for key, value := range headers {
+				req.Header.Add(key, value)
+			}
+			return roundTripper.RoundTrip(req)
+		})
+	}
+
+	// Create new client.
+	client, err := api.NewClient(config)
+	if err != nil {
+		return nil, err
+	}
+
+	api := v1.NewAPI(client)
+	return api, nil
+}
+
+// QueryInstant performs an instant query against a Prometheus server.
+func QueryInstant(url *url.URL, roundTripper http.RoundTripper, query, evalTime string, p printer) int {
+	api, err := newAPI(url, roundTripper, nil)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "error creating API client:", err)
+		return failureExitCode
+	}
+
+	eTime := time.Now()
+	if evalTime != "" {
+		eTime, err = parseTime(evalTime)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "error parsing evaluation time:", err)
+			return failureExitCode
+		}
+	}
+
+	// Run query against client.
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
+	val, _, err := api.Query(ctx, query, eTime) // Ignoring warnings for now.
+	cancel()
+	if err != nil {
+		return handleAPIError(err)
+	}
+
+	p.printValue(val)
+
+	return successExitCode
+}
+
+// QueryRange performs a range query against a Prometheus server.
+func QueryRange(url *url.URL, roundTripper http.RoundTripper, headers map[string]string, query, start, end string, step time.Duration, p printer) int {
+	api, err := newAPI(url, roundTripper, headers)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "error creating API client:", err)
+		return failureExitCode
+	}
+
+	var stime, etime time.Time
+
+	if end == "" {
+		etime = time.Now()
+	} else {
+		etime, err = parseTime(end)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "error parsing end time:", err)
+			return failureExitCode
+		}
+	}
+
+	if start == "" {
+		stime = etime.Add(-5 * time.Minute)
+	} else {
+		stime, err = parseTime(start)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "error parsing start time:", err)
+			return failureExitCode
+		}
+	}
+
+	if !stime.Before(etime) {
+		fmt.Fprintln(os.Stderr, "start time is not before end time")
+		return failureExitCode
+	}
+
+	if step == 0 {
+		resolution := math.Max(math.Floor(etime.Sub(stime).Seconds()/250), 1)
+		// Convert seconds to nanoseconds such that time.Duration parses correctly.
+		step = time.Duration(resolution) * time.Second
+	}
+
+	// Run query against client.
+	r := v1.Range{Start: stime, End: etime, Step: step}
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
+	val, _, err := api.QueryRange(ctx, query, r) // Ignoring warnings for now.
+	cancel()
+
+	if err != nil {
+		return handleAPIError(err)
+	}
+
+	p.printValue(val)
+	return successExitCode
+}
+
+// QuerySeries queries for a series against a Prometheus server.
+func QuerySeries(url *url.URL, roundTripper http.RoundTripper, matchers []string, start, end string, p printer) int {
+	api, err := newAPI(url, roundTripper, nil)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "error creating API client:", err)
+		return failureExitCode
+	}
+
+	stime, etime, err := parseStartTimeAndEndTime(start, end)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return failureExitCode
+	}
+
+	// Run query against client.
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
+	val, _, err := api.Series(ctx, matchers, stime, etime) // Ignoring warnings for now.
+	cancel()
+
+	if err != nil {
+		return handleAPIError(err)
+	}
+
+	p.printSeries(val)
+	return successExitCode
+}
+
+// QueryLabels queries for label values against a Prometheus server.
+func QueryLabels(url *url.URL, roundTripper http.RoundTripper, matchers []string, name, start, end string, p printer) int {
+	api, err := newAPI(url, roundTripper, nil)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, "error creating API client:", err)
+		return failureExitCode
+	}
+
+	stime, etime, err := parseStartTimeAndEndTime(start, end)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return failureExitCode
+	}
+
+	// Run query against client.
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
+	val, warn, err := api.LabelValues(ctx, name, matchers, stime, etime)
+	cancel()
+
+	for _, v := range warn {
+		fmt.Fprintln(os.Stderr, "query warning:", v)
+	}
+	if err != nil {
+		return handleAPIError(err)
+	}
+
+	p.printLabelValues(val)
+	return successExitCode
+}
+
+func handleAPIError(err error) int {
+	var apiErr *v1.Error
+	if errors.As(err, &apiErr) && apiErr.Detail != "" {
+		fmt.Fprintf(os.Stderr, "query error: %v (detail: %s)\n", apiErr, strings.TrimSpace(apiErr.Detail))
+	} else {
+		fmt.Fprintln(os.Stderr, "query error:", err)
+	}
+
+	return failureExitCode
+}
+
+func parseStartTimeAndEndTime(start, end string) (time.Time, time.Time, error) {
+	var (
+		minTime = time.Now().Add(-9999 * time.Hour)
+		maxTime = time.Now().Add(9999 * time.Hour)
+		err     error
+	)
+
+	stime := minTime
+	etime := maxTime
+
+	if start != "" {
+		stime, err = parseTime(start)
+		if err != nil {
+			return stime, etime, fmt.Errorf("error parsing start time: %w", err)
+		}
+	}
+
+	if end != "" {
+		etime, err = parseTime(end)
+		if err != nil {
+			return stime, etime, fmt.Errorf("error parsing end time: %w", err)
+		}
+	}
+	return stime, etime, nil
+}
+
+func parseTime(s string) (time.Time, error) {
+	if t, err := strconv.ParseFloat(s, 64); err == nil {
+		s, ns := math.Modf(t)
+		return time.Unix(int64(s), int64(ns*float64(time.Second))).UTC(), nil
+	}
+	if t, err := time.Parse(time.RFC3339Nano, s); err == nil {
+		return t, nil
+	}
+	return time.Time{}, fmt.Errorf("cannot parse %q to a valid timestamp", s)
+}

cmd/promtool/rules.go

@@ -234,17 +234,3 @@ func (m *multipleAppender) flushAndCommit(ctx context.Context) error {
 	}
 	return nil
 }
-
-func max(x, y int64) int64 {
-	if x > y {
-		return x
-	}
-	return y
-}
-
-func min(x, y int64) int64 {
-	if x < y {
-		return x
-	}
-	return y
-}

cmd/promtool/rules_test.go

@@ -78,7 +78,6 @@ func TestBackfillRuleIntegration(t *testing.T) {
 			// Execute the test more than once to simulate running the rule importer twice with the same data.
 			// We expect duplicate blocks with the same series are created when run more than once.
 			for i := 0; i < tt.runcount; i++ {
-
 				ruleImporter, err := newTestRuleImporter(ctx, start, tmpDir, tt.samples, tt.maxBlockDuration)
 				require.NoError(t, err)
 				path1 := filepath.Join(tmpDir, "test.file")

cmd/promtool/sd.go

@@ -18,10 +18,10 @@ import (
 	"encoding/json"
 	"fmt"
 	"os"
-	"reflect"
 	"time"
 
 	"github.com/go-kit/log"
+	"github.com/google/go-cmp/cmp"
 	"github.com/prometheus/client_golang/prometheus"
 
 	"github.com/prometheus/prometheus/config"
@@ -78,12 +78,25 @@ func CheckSD(sdConfigFiles, sdJobName string, sdTimeout time.Duration, noDefault
 	defer cancel()
 
 	for _, cfg := range scrapeConfig.ServiceDiscoveryConfigs {
-		d, err := cfg.NewDiscoverer(discovery.DiscovererOptions{Logger: logger, Registerer: registerer})
+		reg := prometheus.NewRegistry()
+		refreshMetrics := discovery.NewRefreshMetrics(reg)
+		metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+		err := metrics.Register()
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "Could not register service discovery metrics", err)
+			return failureExitCode
+		}
+
+		d, err := cfg.NewDiscoverer(discovery.DiscovererOptions{Logger: logger, Metrics: metrics})
 		if err != nil {
 			fmt.Fprintln(os.Stderr, "Could not create new discoverer", err)
 			return failureExitCode
 		}
-		go d.Run(ctx, targetGroupChan)
+		go func() {
+			d.Run(ctx, targetGroupChan)
+			metrics.Unregister()
+			refreshMetrics.Unregister()
+		}()
 	}
 
 	var targetGroups []*targetgroup.Group
@@ -140,7 +153,7 @@ func getSDCheckResult(targetGroups []*targetgroup.Group, scrapeConfig *config.Sc
 
 			duplicateRes := false
 			for _, sdCheckRes := range sdCheckResults {
-				if reflect.DeepEqual(sdCheckRes, result) {
+				if cmp.Equal(sdCheckRes, result, cmp.Comparer(labels.Equal)) {
 					duplicateRes = true
 					break
 				}

cmd/promtool/sd_test.go

@@ -23,6 +23,7 @@ import (
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 	"github.com/prometheus/prometheus/model/labels"
 	"github.com/prometheus/prometheus/model/relabel"
+	"github.com/prometheus/prometheus/util/testutil"
 
 	"github.com/stretchr/testify/require"
 )
@@ -69,5 +70,5 @@ func TestSDCheckResult(t *testing.T) {
 		},
 	}
 
-	require.Equal(t, expectedSDCheckResult, getSDCheckResult(targetGroups, scrapeConfig, true))
+	testutil.RequireEqual(t, expectedSDCheckResult, getSDCheckResult(targetGroups, scrapeConfig, true))
 }

cmd/promtool/testdata/dump-openmetrics-roundtrip-test.prom

@@ -0,0 +1,15 @@
+my_histogram_bucket{instance="localhost:8000",job="example2",le="+Inf"} 1.0267820369e+10 1700215884.373
+my_histogram_bucket{instance="localhost:8000",job="example2",le="+Inf"} 1.026872507e+10 1700215889.373
+my_histogram_bucket{instance="localhost:8000",job="example2",le="0.01"} 0 1700215884.373
+my_histogram_bucket{instance="localhost:8000",job="example2",le="0.01"} 0 1700215889.373
+my_histogram_count{instance="localhost:8000",job="example2"} 1.0267820369e+10 1700215884.373
+my_histogram_count{instance="localhost:8000",job="example2"} 1.026872507e+10 1700215889.373
+my_summary_count{instance="localhost:8000",job="example5"} 9.518161497e+09 1700211684.981
+my_summary_count{instance="localhost:8000",job="example5"} 9.519048034e+09 1700211689.984
+my_summary_sum{instance="localhost:8000",job="example5"} 5.2349889185e+10 1700211684.981
+my_summary_sum{instance="localhost:8000",job="example5"} 5.2354761848e+10 1700211689.984
+up{instance="localhost:8000",job="example2"} 1 1700226034.330
+up{instance="localhost:8000",job="example2"} 1 1700226094.329
+up{instance="localhost:8000",job="example3"} 1 1700210681.366
+up{instance="localhost:8000",job="example3"} 1 1700210686.366
+# EOF

cmd/promtool/testdata/dump-openmetrics-test.prom

@@ -0,0 +1,11 @@
+my_counter{baz="abc",foo="bar"} 1 0.000
+my_counter{baz="abc",foo="bar"} 2 60.000
+my_counter{baz="abc",foo="bar"} 3 120.000
+my_counter{baz="abc",foo="bar"} 4 180.000
+my_counter{baz="abc",foo="bar"} 5 240.000
+my_gauge{abc="baz",bar="foo"} 9 0.000
+my_gauge{abc="baz",bar="foo"} 8 60.000
+my_gauge{abc="baz",bar="foo"} 0 120.000
+my_gauge{abc="baz",bar="foo"} 4 180.000
+my_gauge{abc="baz",bar="foo"} 7 240.000
+# EOF

cmd/promtool/testdata/dump-test-1.prom

@@ -0,0 +1,15 @@
+{__name__="heavy_metric", foo="bar"} 5 0
+{__name__="heavy_metric", foo="bar"} 4 60000
+{__name__="heavy_metric", foo="bar"} 3 120000
+{__name__="heavy_metric", foo="bar"} 2 180000
+{__name__="heavy_metric", foo="bar"} 1 240000
+{__name__="heavy_metric", foo="foo"} 5 0
+{__name__="heavy_metric", foo="foo"} 4 60000
+{__name__="heavy_metric", foo="foo"} 3 120000
+{__name__="heavy_metric", foo="foo"} 2 180000
+{__name__="heavy_metric", foo="foo"} 1 240000
+{__name__="metric", baz="abc", foo="bar"} 1 0
+{__name__="metric", baz="abc", foo="bar"} 2 60000
+{__name__="metric", baz="abc", foo="bar"} 3 120000
+{__name__="metric", baz="abc", foo="bar"} 4 180000
+{__name__="metric", baz="abc", foo="bar"} 5 240000

cmd/promtool/testdata/dump-test-2.prom

@@ -0,0 +1,10 @@
+{__name__="heavy_metric", foo="foo"} 5 0
+{__name__="heavy_metric", foo="foo"} 4 60000
+{__name__="heavy_metric", foo="foo"} 3 120000
+{__name__="heavy_metric", foo="foo"} 2 180000
+{__name__="heavy_metric", foo="foo"} 1 240000
+{__name__="metric", baz="abc", foo="bar"} 1 0
+{__name__="metric", baz="abc", foo="bar"} 2 60000
+{__name__="metric", baz="abc", foo="bar"} 3 120000
+{__name__="metric", baz="abc", foo="bar"} 4 180000
+{__name__="metric", baz="abc", foo="bar"} 5 240000

cmd/promtool/testdata/dump-test-3.prom

@@ -0,0 +1,2 @@
+{__name__="metric", baz="abc", foo="bar"} 2 60000
+{__name__="metric", baz="abc", foo="bar"} 3 120000

cmd/promtool/testdata/no-test-group-interval.yml

@@ -12,4 +12,4 @@ tests:
         eval_time: 1m
         exp_samples:
           - value: 1
-            labels: test
+            labels: test

cmd/promtool/tsdb.go

@@ -15,6 +15,7 @@ package main
 
 import (
 	"bufio"
+	"bytes"
 	"context"
 	"errors"
 	"fmt"
@@ -23,6 +24,7 @@ import (
 	"path/filepath"
 	"runtime"
 	"runtime/pprof"
+	"slices"
 	"strconv"
 	"strings"
 	"sync"
@@ -31,7 +33,7 @@ import (
 
 	"github.com/alecthomas/units"
 	"github.com/go-kit/log"
-	"golang.org/x/exp/slices"
+	"go.uber.org/atomic"
 
 	"github.com/prometheus/prometheus/model/labels"
 	"github.com/prometheus/prometheus/promql/parser"
@@ -148,8 +150,7 @@ func benchmarkWrite(outPath, samplesFile string, numMetrics, numScrapes int) err
 }
 
 func (b *writeBenchmark) ingestScrapes(lbls []labels.Labels, scrapeCount int) (uint64, error) {
-	var mu sync.Mutex
-	var total uint64
+	var total atomic.Uint64
 
 	for i := 0; i < scrapeCount; i += 100 {
 		var wg sync.WaitGroup
@@ -164,22 +165,21 @@ func (b *writeBenchmark) ingestScrapes(lbls []labels.Labels, scrapeCount int) (u
 
 			wg.Add(1)
 			go func() {
+				defer wg.Done()
+
 				n, err := b.ingestScrapesShard(batch, 100, int64(timeDelta*i))
 				if err != nil {
 					// exitWithError(err)
 					fmt.Println(" err", err)
 				}
-				mu.Lock()
-				total += n
-				mu.Unlock()
-				wg.Done()
+				total.Add(n)
 			}()
 		}
 		wg.Wait()
 	}
 	fmt.Println("ingestion completed")
 
-	return total, nil
+	return total.Load(), nil
 }
 
 func (b *writeBenchmark) ingestScrapesShard(lbls []labels.Labels, scrapeCount int, baset int64) (uint64, error) {
@@ -338,7 +338,7 @@ func readPrometheusLabels(r io.Reader, n int) ([]labels.Labels, error) {
 }
 
 func listBlocks(path string, humanReadable bool) error {
-	db, err := tsdb.OpenDBReadOnly(path, nil)
+	db, err := tsdb.OpenDBReadOnly(path, "", nil)
 	if err != nil {
 		return err
 	}
@@ -393,7 +393,7 @@ func getFormatedBytes(bytes int64, humanReadable bool) string {
 }
 
 func openBlock(path, blockID string) (*tsdb.DBReadOnly, tsdb.BlockReader, error) {
-	db, err := tsdb.OpenDBReadOnly(path, nil)
+	db, err := tsdb.OpenDBReadOnly(path, "", nil)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -667,7 +667,7 @@ func analyzeCompaction(ctx context.Context, block tsdb.BlockReader, indexr tsdb.
 				it := fhchk.Iterator(nil)
 				bucketCount := 0
 				for it.Next() == chunkenc.ValFloatHistogram {
-					_, f := it.AtFloatHistogram()
+					_, f := it.AtFloatHistogram(nil)
 					bucketCount += len(f.PositiveBuckets)
 					bucketCount += len(f.NegativeBuckets)
 				}
@@ -682,7 +682,7 @@ func analyzeCompaction(ctx context.Context, block tsdb.BlockReader, indexr tsdb.
 				it := hchk.Iterator(nil)
 				bucketCount := 0
 				for it.Next() == chunkenc.ValHistogram {
-					_, f := it.AtHistogram()
+					_, f := it.AtHistogram(nil)
 					bucketCount += len(f.PositiveBuckets)
 					bucketCount += len(f.NegativeBuckets)
 				}
@@ -706,8 +706,10 @@ func analyzeCompaction(ctx context.Context, block tsdb.BlockReader, indexr tsdb.
 	return nil
 }
 
-func dumpSamples(ctx context.Context, path string, mint, maxt int64, match string) (err error) {
-	db, err := tsdb.OpenDBReadOnly(path, nil)
+type SeriesSetFormatter func(series storage.SeriesSet) error
+
+func dumpSamples(ctx context.Context, dbDir, sandboxDirRoot string, mint, maxt int64, match []string, formatter SeriesSetFormatter) (err error) {
+	db, err := tsdb.OpenDBReadOnly(dbDir, sandboxDirRoot, nil)
 	if err != nil {
 		return err
 	}
@@ -720,31 +722,25 @@ func dumpSamples(ctx context.Context, path string, mint, maxt int64, match strin
 	}
 	defer q.Close()
 
-	matchers, err := parser.ParseMetricSelector(match)
+	matcherSets, err := parser.ParseMetricSelectors(match)
 	if err != nil {
 		return err
 	}
-	ss := q.Select(ctx, false, nil, matchers...)
 
-	for ss.Next() {
-		series := ss.At()
-		lbs := series.Labels()
-		it := series.Iterator(nil)
-		for it.Next() == chunkenc.ValFloat {
-			ts, val := it.At()
-			fmt.Printf("%s %g %d\n", lbs, val, ts)
-		}
-		for it.Next() == chunkenc.ValFloatHistogram {
-			ts, fh := it.AtFloatHistogram()
-			fmt.Printf("%s %s %d\n", lbs, fh.String(), ts)
-		}
-		for it.Next() == chunkenc.ValHistogram {
-			ts, h := it.AtHistogram()
-			fmt.Printf("%s %s %d\n", lbs, h.String(), ts)
-		}
-		if it.Err() != nil {
-			return ss.Err()
+	var ss storage.SeriesSet
+	if len(matcherSets) > 1 {
+		var sets []storage.SeriesSet
+		for _, mset := range matcherSets {
+			sets = append(sets, q.Select(ctx, true, nil, mset...))
 		}
+		ss = storage.NewMergeSeriesSet(sets, storage.ChainedSeriesMerge)
+	} else {
+		ss = q.Select(ctx, false, nil, matcherSets[0]...)
+	}
+
+	err = formatter(ss)
+	if err != nil {
+		return err
 	}
 
 	if ws := ss.Warnings(); len(ws) > 0 {
@@ -757,6 +753,68 @@ func dumpSamples(ctx context.Context, path string, mint, maxt int64, match strin
 	return nil
 }
 
+func formatSeriesSet(ss storage.SeriesSet) error {
+	for ss.Next() {
+		series := ss.At()
+		lbs := series.Labels()
+		it := series.Iterator(nil)
+		for it.Next() == chunkenc.ValFloat {
+			ts, val := it.At()
+			fmt.Printf("%s %g %d\n", lbs, val, ts)
+		}
+		for it.Next() == chunkenc.ValFloatHistogram {
+			ts, fh := it.AtFloatHistogram(nil)
+			fmt.Printf("%s %s %d\n", lbs, fh.String(), ts)
+		}
+		for it.Next() == chunkenc.ValHistogram {
+			ts, h := it.AtHistogram(nil)
+			fmt.Printf("%s %s %d\n", lbs, h.String(), ts)
+		}
+		if it.Err() != nil {
+			return ss.Err()
+		}
+	}
+	return nil
+}
+
+// CondensedString is labels.Labels.String() without spaces after the commas.
+func CondensedString(ls labels.Labels) string {
+	var b bytes.Buffer
+
+	b.WriteByte('{')
+	i := 0
+	ls.Range(func(l labels.Label) {
+		if i > 0 {
+			b.WriteByte(',')
+		}
+		b.WriteString(l.Name)
+		b.WriteByte('=')
+		b.WriteString(strconv.Quote(l.Value))
+		i++
+	})
+	b.WriteByte('}')
+	return b.String()
+}
+
+func formatSeriesSetOpenMetrics(ss storage.SeriesSet) error {
+	for ss.Next() {
+		series := ss.At()
+		lbs := series.Labels()
+		metricName := lbs.Get(labels.MetricName)
+		lbs = lbs.DropMetricName()
+		it := series.Iterator(nil)
+		for it.Next() == chunkenc.ValFloat {
+			ts, val := it.At()
+			fmt.Printf("%s%s %g %.3f\n", metricName, CondensedString(lbs), val, float64(ts)/1000)
+		}
+		if it.Err() != nil {
+			return ss.Err()
+		}
+	}
+	fmt.Println("# EOF")
+	return nil
+}
+
 func checkErr(err error) int {
 	if err != nil {
 		fmt.Fprintln(os.Stderr, err)
@@ -780,6 +838,10 @@ func backfillOpenMetrics(path, outputDir string, humanReadable, quiet bool, maxB
 }
 
 func displayHistogram(dataType string, datas []int, total int) {
+	if len(datas) == 0 {
+		fmt.Printf("%s: N/A\n\n", dataType)
+		return
+	}
 	slices.Sort(datas)
 	start, end, step := generateBucket(datas[0], datas[len(datas)-1])
 	sum := 0
@@ -794,9 +856,9 @@ func displayHistogram(dataType string, datas []int, total int) {
 	}
 	avg := sum / len(datas)
 	fmt.Printf("%s (min/avg/max): %d/%d/%d\n", dataType, datas[0], avg, datas[len(datas)-1])
-	maxLeftLen := strconv.Itoa(len(fmt.Sprintf("%d", end)))
-	maxRightLen := strconv.Itoa(len(fmt.Sprintf("%d", end+step)))
-	maxCountLen := strconv.Itoa(len(fmt.Sprintf("%d", maxCount)))
+	maxLeftLen := strconv.Itoa(len(strconv.Itoa(end)))
+	maxRightLen := strconv.Itoa(len(strconv.Itoa(end + step)))
+	maxCountLen := strconv.Itoa(len(strconv.Itoa(maxCount)))
 	for bucket, count := range buckets {
 		percentage := 100.0 * count / total
 		fmt.Printf("[%"+maxLeftLen+"d, %"+maxRightLen+"d]: %"+maxCountLen+"d %s\n", bucket*step+start+1, (bucket+1)*step+start, count, strings.Repeat("#", percentage))

cmd/promtool/tsdb_test.go

@@ -14,9 +14,20 @@
 package main
 
 import (
+	"bytes"
+	"context"
+	"io"
+	"math"
+	"os"
+	"runtime"
+	"strings"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/require"
+
+	"github.com/prometheus/prometheus/promql/promqltest"
+	"github.com/prometheus/prometheus/tsdb"
 )
 
 func TestGenerateBucket(t *testing.T) {
@@ -41,3 +52,144 @@ func TestGenerateBucket(t *testing.T) {
 		require.Equal(t, tc.step, step)
 	}
 }
+
+// getDumpedSamples dumps samples and returns them.
+func getDumpedSamples(t *testing.T, path string, mint, maxt int64, match []string, formatter SeriesSetFormatter) string {
+	t.Helper()
+
+	oldStdout := os.Stdout
+	r, w, _ := os.Pipe()
+	os.Stdout = w
+
+	err := dumpSamples(
+		context.Background(),
+		path,
+		t.TempDir(),
+		mint,
+		maxt,
+		match,
+		formatter,
+	)
+	require.NoError(t, err)
+
+	w.Close()
+	os.Stdout = oldStdout
+
+	var buf bytes.Buffer
+	io.Copy(&buf, r)
+	return buf.String()
+}
+
+func normalizeNewLine(b []byte) []byte {
+	if strings.Contains(runtime.GOOS, "windows") {
+		// We use "/n" while dumping on windows as well.
+		return bytes.ReplaceAll(b, []byte("\r\n"), []byte("\n"))
+	}
+	return b
+}
+
+func TestTSDBDump(t *testing.T) {
+	storage := promqltest.LoadedStorage(t, `
+		load 1m
+			metric{foo="bar", baz="abc"} 1 2 3 4 5
+			heavy_metric{foo="bar"} 5 4 3 2 1
+			heavy_metric{foo="foo"} 5 4 3 2 1
+	`)
+
+	tests := []struct {
+		name         string
+		mint         int64
+		maxt         int64
+		match        []string
+		expectedDump string
+	}{
+		{
+			name:         "default match",
+			mint:         math.MinInt64,
+			maxt:         math.MaxInt64,
+			match:        []string{"{__name__=~'(?s:.*)'}"},
+			expectedDump: "testdata/dump-test-1.prom",
+		},
+		{
+			name:         "same matcher twice",
+			mint:         math.MinInt64,
+			maxt:         math.MaxInt64,
+			match:        []string{"{foo=~'.+'}", "{foo=~'.+'}"},
+			expectedDump: "testdata/dump-test-1.prom",
+		},
+		{
+			name:         "no duplication",
+			mint:         math.MinInt64,
+			maxt:         math.MaxInt64,
+			match:        []string{"{__name__=~'(?s:.*)'}", "{baz='abc'}"},
+			expectedDump: "testdata/dump-test-1.prom",
+		},
+		{
+			name:         "well merged",
+			mint:         math.MinInt64,
+			maxt:         math.MaxInt64,
+			match:        []string{"{__name__='heavy_metric'}", "{baz='abc'}"},
+			expectedDump: "testdata/dump-test-1.prom",
+		},
+		{
+			name:         "multi matchers",
+			mint:         math.MinInt64,
+			maxt:         math.MaxInt64,
+			match:        []string{"{__name__='heavy_metric',foo='foo'}", "{__name__='metric'}"},
+			expectedDump: "testdata/dump-test-2.prom",
+		},
+		{
+			name:         "with reduced mint and maxt",
+			mint:         int64(60000),
+			maxt:         int64(120000),
+			match:        []string{"{__name__='metric'}"},
+			expectedDump: "testdata/dump-test-3.prom",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			dumpedMetrics := getDumpedSamples(t, storage.Dir(), tt.mint, tt.maxt, tt.match, formatSeriesSet)
+			expectedMetrics, err := os.ReadFile(tt.expectedDump)
+			require.NoError(t, err)
+			expectedMetrics = normalizeNewLine(expectedMetrics)
+			// even though in case of one matcher samples are not sorted, the order in the cases above should stay the same.
+			require.Equal(t, string(expectedMetrics), dumpedMetrics)
+		})
+	}
+}
+
+func TestTSDBDumpOpenMetrics(t *testing.T) {
+	storage := promqltest.LoadedStorage(t, `
+		load 1m
+			my_counter{foo="bar", baz="abc"} 1 2 3 4 5
+			my_gauge{bar="foo", abc="baz"} 9 8 0 4 7
+	`)
+
+	expectedMetrics, err := os.ReadFile("testdata/dump-openmetrics-test.prom")
+	require.NoError(t, err)
+	expectedMetrics = normalizeNewLine(expectedMetrics)
+	dumpedMetrics := getDumpedSamples(t, storage.Dir(), math.MinInt64, math.MaxInt64, []string{"{__name__=~'(?s:.*)'}"}, formatSeriesSetOpenMetrics)
+	require.Equal(t, string(expectedMetrics), dumpedMetrics)
+}
+
+func TestTSDBDumpOpenMetricsRoundTrip(t *testing.T) {
+	initialMetrics, err := os.ReadFile("testdata/dump-openmetrics-roundtrip-test.prom")
+	require.NoError(t, err)
+	initialMetrics = normalizeNewLine(initialMetrics)
+
+	dbDir := t.TempDir()
+	// Import samples from OM format
+	err = backfill(5000, initialMetrics, dbDir, false, false, 2*time.Hour)
+	require.NoError(t, err)
+	db, err := tsdb.Open(dbDir, nil, nil, tsdb.DefaultOptions(), nil)
+	require.NoError(t, err)
+	t.Cleanup(func() {
+		require.NoError(t, db.Close())
+	})
+
+	// Dump the blocks into OM format
+	dumpedMetrics := getDumpedSamples(t, dbDir, math.MinInt64, math.MaxInt64, []string{"{__name__=~'(?s:.*)'}"}, formatSeriesSetOpenMetrics)
+
+	// Should get back the initial metrics.
+	require.Equal(t, string(initialMetrics), dumpedMetrics)
+}

cmd/promtool/unittest.go

@@ -15,18 +15,20 @@ package main
 
 import (
 	"context"
+	"encoding/json"
 	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
-	"reflect"
 	"sort"
 	"strconv"
 	"strings"
 	"time"
 
 	"github.com/go-kit/log"
+	"github.com/google/go-cmp/cmp"
 	"github.com/grafana/regexp"
+	"github.com/nsf/jsondiff"
 	"github.com/prometheus/common/model"
 	"gopkg.in/yaml.v2"
 
@@ -34,13 +36,14 @@ import (
 	"github.com/prometheus/prometheus/model/labels"
 	"github.com/prometheus/prometheus/promql"
 	"github.com/prometheus/prometheus/promql/parser"
+	"github.com/prometheus/prometheus/promql/promqltest"
 	"github.com/prometheus/prometheus/rules"
 	"github.com/prometheus/prometheus/storage"
 )
 
 // RulesUnitTest does unit testing of rules based on the unit testing files provided.
 // More info about the file format can be found in the docs.
-func RulesUnitTest(queryOpts promql.LazyLoaderOpts, runStrings []string, files ...string) int {
+func RulesUnitTest(queryOpts promqltest.LazyLoaderOpts, runStrings []string, diffFlag bool, files ...string) int {
 	failed := false
 
 	var run *regexp.Regexp
@@ -49,7 +52,7 @@ func RulesUnitTest(queryOpts promql.LazyLoaderOpts, runStrings []string, files .
 	}
 
 	for _, f := range files {
-		if errs := ruleUnitTest(f, queryOpts, run); errs != nil {
+		if errs := ruleUnitTest(f, queryOpts, run, diffFlag); errs != nil {
 			fmt.Fprintln(os.Stderr, "  FAILED:")
 			for _, e := range errs {
 				fmt.Fprintln(os.Stderr, e.Error())
@@ -67,7 +70,7 @@ func RulesUnitTest(queryOpts promql.LazyLoaderOpts, runStrings []string, files .
 	return successExitCode
 }
 
-func ruleUnitTest(filename string, queryOpts promql.LazyLoaderOpts, run *regexp.Regexp) []error {
+func ruleUnitTest(filename string, queryOpts promqltest.LazyLoaderOpts, run *regexp.Regexp, diffFlag bool) []error {
 	fmt.Println("Unit Testing: ", filename)
 
 	b, err := os.ReadFile(filename)
@@ -109,7 +112,7 @@ func ruleUnitTest(filename string, queryOpts promql.LazyLoaderOpts, run *regexp.
 		if t.Interval == 0 {
 			t.Interval = unitTestInp.EvaluationInterval
 		}
-		ers := t.test(evalInterval, groupOrderMap, queryOpts, unitTestInp.RuleFiles...)
+		ers := t.test(evalInterval, groupOrderMap, queryOpts, diffFlag, unitTestInp.RuleFiles...)
 		if ers != nil {
 			errs = append(errs, ers...)
 		}
@@ -173,13 +176,18 @@ type testGroup struct {
 }
 
 // test performs the unit tests.
-func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]int, queryOpts promql.LazyLoaderOpts, ruleFiles ...string) []error {
+func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]int, queryOpts promqltest.LazyLoaderOpts, diffFlag bool, ruleFiles ...string) (outErr []error) {
 	// Setup testing suite.
-	suite, err := promql.NewLazyLoader(nil, tg.seriesLoadingString(), queryOpts)
+	suite, err := promqltest.NewLazyLoader(tg.seriesLoadingString(), queryOpts)
 	if err != nil {
 		return []error{err}
 	}
-	defer suite.Close()
+	defer func() {
+		err := suite.Close()
+		if err != nil {
+			outErr = append(outErr, err)
+		}
+	}()
 	suite.SubqueryInterval = evalInterval
 
 	// Load the rule files.
@@ -338,15 +346,51 @@ func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]i
 				sort.Sort(gotAlerts)
 				sort.Sort(expAlerts)
 
-				if !reflect.DeepEqual(expAlerts, gotAlerts) {
+				if !cmp.Equal(expAlerts, gotAlerts, cmp.Comparer(labels.Equal)) {
 					var testName string
 					if tg.TestGroupName != "" {
 						testName = fmt.Sprintf("    name: %s,\n", tg.TestGroupName)
 					}
 					expString := indentLines(expAlerts.String(), "            ")
 					gotString := indentLines(gotAlerts.String(), "            ")
-					errs = append(errs, fmt.Errorf("%s    alertname: %s, time: %s, \n        exp:%v, \n        got:%v",
-						testName, testcase.Alertname, testcase.EvalTime.String(), expString, gotString))
+					if diffFlag {
+						// If empty, populates an empty value
+						if gotAlerts.Len() == 0 {
+							gotAlerts = append(gotAlerts, labelAndAnnotation{
+								Labels:      labels.Labels{},
+								Annotations: labels.Labels{},
+							})
+						}
+						// If empty, populates an empty value
+						if expAlerts.Len() == 0 {
+							expAlerts = append(expAlerts, labelAndAnnotation{
+								Labels:      labels.Labels{},
+								Annotations: labels.Labels{},
+							})
+						}
+
+						diffOpts := jsondiff.DefaultConsoleOptions()
+						expAlertsJSON, err := json.Marshal(expAlerts)
+						if err != nil {
+							errs = append(errs, fmt.Errorf("error marshaling expected %s alert: [%s]", tg.TestGroupName, err.Error()))
+							continue
+						}
+
+						gotAlertsJSON, err := json.Marshal(gotAlerts)
+						if err != nil {
+							errs = append(errs, fmt.Errorf("error marshaling received %s alert: [%s]", tg.TestGroupName, err.Error()))
+							continue
+						}
+
+						res, diff := jsondiff.Compare(expAlertsJSON, gotAlertsJSON, &diffOpts)
+						if res != jsondiff.FullMatch {
+							errs = append(errs, fmt.Errorf("%s    alertname: %s, time: %s, \n        diff: %v",
+								testName, testcase.Alertname, testcase.EvalTime.String(), indentLines(diff, "            ")))
+						}
+					} else {
+						errs = append(errs, fmt.Errorf("%s    alertname: %s, time: %s, \n        exp:%v, \n        got:%v",
+							testName, testcase.Alertname, testcase.EvalTime.String(), expString, gotString))
+					}
 				}
 			}
 
@@ -370,7 +414,7 @@ Outer:
 			gotSamples = append(gotSamples, parsedSample{
 				Labels:    s.Metric.Copy(),
 				Value:     s.F,
-				Histogram: promql.HistogramTestExpression(s.H),
+				Histogram: promqltest.HistogramTestExpression(s.H),
 			})
 		}
 
@@ -400,7 +444,7 @@ Outer:
 			expSamples = append(expSamples, parsedSample{
 				Labels:    lb,
 				Value:     s.Value,
-				Histogram: promql.HistogramTestExpression(hist),
+				Histogram: promqltest.HistogramTestExpression(hist),
 			})
 		}
 
@@ -410,7 +454,7 @@ Outer:
 		sort.Slice(gotSamples, func(i, j int) bool {
 			return labels.Compare(gotSamples[i].Labels, gotSamples[j].Labels) <= 0
 		})
-		if !reflect.DeepEqual(expSamples, gotSamples) {
+		if !cmp.Equal(expSamples, gotSamples, cmp.Comparer(labels.Equal)) {
 			errs = append(errs, fmt.Errorf("    expr: %q, time: %s,\n        exp: %v\n        got: %v", testCase.Expr,
 				testCase.EvalTime.String(), parsedSamplesString(expSamples), parsedSamplesString(gotSamples)))
 		}
@@ -529,7 +573,7 @@ func (la labelsAndAnnotations) String() string {
 	}
 	s := "[\n0:" + indentLines("\n"+la[0].String(), "  ")
 	for i, l := range la[1:] {
-		s += ",\n" + fmt.Sprintf("%d", i+1) + ":" + indentLines("\n"+l.String(), "  ")
+		s += ",\n" + strconv.Itoa(i+1) + ":" + indentLines("\n"+l.String(), "  ")
 	}
 	s += "\n]"
 

cmd/promtool/unittest_test.go

@@ -16,7 +16,9 @@ package main
 import (
 	"testing"
 
-	"github.com/prometheus/prometheus/promql"
+	"github.com/stretchr/testify/require"
+
+	"github.com/prometheus/prometheus/promql/promqltest"
 )
 
 func TestRulesUnitTest(t *testing.T) {
@@ -26,7 +28,7 @@ func TestRulesUnitTest(t *testing.T) {
 	tests := []struct {
 		name      string
 		args      args
-		queryOpts promql.LazyLoaderOpts
+		queryOpts promqltest.LazyLoaderOpts
 		want      int
 	}{
 		{
@@ -90,7 +92,7 @@ func TestRulesUnitTest(t *testing.T) {
 			args: args{
 				files: []string{"./testdata/at-modifier-test.yml"},
 			},
-			queryOpts: promql.LazyLoaderOpts{
+			queryOpts: promqltest.LazyLoaderOpts{
 				EnableAtModifier: true,
 			},
 			want: 0,
@@ -107,7 +109,7 @@ func TestRulesUnitTest(t *testing.T) {
 			args: args{
 				files: []string{"./testdata/negative-offset-test.yml"},
 			},
-			queryOpts: promql.LazyLoaderOpts{
+			queryOpts: promqltest.LazyLoaderOpts{
 				EnableNegativeOffset: true,
 			},
 			want: 0,
@@ -117,7 +119,7 @@ func TestRulesUnitTest(t *testing.T) {
 			args: args{
 				files: []string{"./testdata/no-test-group-interval.yml"},
 			},
-			queryOpts: promql.LazyLoaderOpts{
+			queryOpts: promqltest.LazyLoaderOpts{
 				EnableNegativeOffset: true,
 			},
 			want: 0,
@@ -125,7 +127,7 @@ func TestRulesUnitTest(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			if got := RulesUnitTest(tt.queryOpts, nil, tt.args.files...); got != tt.want {
+			if got := RulesUnitTest(tt.queryOpts, nil, false, tt.args.files...); got != tt.want {
 				t.Errorf("RulesUnitTest() = %v, want %v", got, tt.want)
 			}
 		})
@@ -140,7 +142,7 @@ func TestRulesUnitTestRun(t *testing.T) {
 	tests := []struct {
 		name      string
 		args      args
-		queryOpts promql.LazyLoaderOpts
+		queryOpts promqltest.LazyLoaderOpts
 		want      int
 	}{
 		{
@@ -178,9 +180,8 @@ func TestRulesUnitTestRun(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			if got := RulesUnitTest(tt.queryOpts, tt.args.run, tt.args.files...); got != tt.want {
-				t.Errorf("RulesUnitTest() = %v, want %v", got, tt.want)
-			}
+			got := RulesUnitTest(tt.queryOpts, tt.args.run, false, tt.args.files...)
+			require.Equal(t, tt.want, got)
 		})
 	}
 }

config/config.go

@@ -82,7 +82,7 @@ func Load(s string, expandExternalLabels bool, logger log.Logger) (*Config, erro
 		return cfg, nil
 	}
 
-	b := labels.ScratchBuilder{}
+	b := labels.NewScratchBuilder(0)
 	cfg.GlobalConfig.ExternalLabels.Range(func(v labels.Label) {
 		newV := os.Expand(v.Value, func(s string) string {
 			if s == "$" {
@@ -97,6 +97,7 @@ func Load(s string, expandExternalLabels bool, logger log.Logger) (*Config, erro
 		if newV != v.Value {
 			level.Debug(logger).Log("msg", "External label replaced", "label", v.Name, "input", v.Value, "output", newV)
 		}
+		// Note newV can be blank. https://github.com/prometheus/prometheus/issues/11024
 		b.Add(v.Name, newV)
 	})
 	cfg.GlobalConfig.ExternalLabels = b.Labels()
@@ -610,9 +611,12 @@ type ScrapeConfig struct {
 	// More than this label value length post metric-relabeling will cause the
 	// scrape to fail. 0 means no limit.
 	LabelValueLengthLimit uint `yaml:"label_value_length_limit,omitempty"`
-	// More than this many buckets in a native histogram will cause the scrape to
-	// fail.
+	// If there are more than this many buckets in a native histogram,
+	// buckets will be merged to stay within the limit.
 	NativeHistogramBucketLimit uint `yaml:"native_histogram_bucket_limit,omitempty"`
+	// If the growth factor of one bucket to the next is smaller than this,
+	// buckets will be merged to increase the factor sufficiently.
+	NativeHistogramMinBucketFactor float64 `yaml:"native_histogram_min_bucket_factor,omitempty"`
 	// Keep no more than this many dropped targets per job.
 	// 0 means no limit.
 	KeepDroppedTargets uint `yaml:"keep_dropped_targets,omitempty"`
@@ -937,6 +941,8 @@ type AlertmanagerConfig struct {
 
 	// List of Alertmanager relabel configurations.
 	RelabelConfigs []*relabel.Config `yaml:"relabel_configs,omitempty"`
+	// Relabel alerts before sending to the specific alertmanager.
+	AlertRelabelConfigs []*relabel.Config `yaml:"alert_relabel_configs,omitempty"`
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -979,6 +985,12 @@ func (c *AlertmanagerConfig) UnmarshalYAML(unmarshal func(interface{}) error) er
 		}
 	}
 
+	for _, rlcfg := range c.AlertRelabelConfigs {
+		if rlcfg == nil {
+			return errors.New("empty or null Alertmanager alert relabeling rule")
+		}
+	}
+
 	return nil
 }
 
@@ -1124,6 +1136,9 @@ type QueueConfig struct {
 	MinBackoff       model.Duration `yaml:"min_backoff,omitempty"`
 	MaxBackoff       model.Duration `yaml:"max_backoff,omitempty"`
 	RetryOnRateLimit bool           `yaml:"retry_on_http_429,omitempty"`
+
+	// Samples older than the limit will be dropped.
+	SampleAgeLimit model.Duration `yaml:"sample_age_limit,omitempty"`
 }
 
 // MetadataConfig is the configuration for sending metadata to remote

config/config_default_test.go

@@ -12,7 +12,6 @@
 // limitations under the License.
 
 //go:build !windows
-// +build !windows
 
 package config
 

config/config_test.go

@@ -58,6 +58,7 @@ import (
 	"github.com/prometheus/prometheus/discovery/zookeeper"
 	"github.com/prometheus/prometheus/model/labels"
 	"github.com/prometheus/prometheus/model/relabel"
+	"github.com/prometheus/prometheus/util/testutil"
 )
 
 func mustParseURL(u string) *config.URL {
@@ -568,6 +569,7 @@ var expectedConf = &Config{
 			ServiceDiscoveryConfigs: discovery.Configs{
 				&xds.KumaSDConfig{
 					Server:           "http://kuma-control-plane.kuma-system.svc:5676",
+					ClientID:         "main-prometheus",
 					HTTPClientConfig: config.DefaultHTTPClientConfig,
 					RefreshInterval:  model.Duration(15 * time.Second),
 					FetchTimeout:     model.Duration(2 * time.Minute),
@@ -1838,7 +1840,7 @@ var expectedErrors = []struct {
 	},
 	{
 		filename: "azure_authentication_method.bad.yml",
-		errMsg:   "unknown authentication_type \"invalid\". Supported types are \"OAuth\" or \"ManagedIdentity\"",
+		errMsg:   "unknown authentication_type \"invalid\". Supported types are \"OAuth\", \"ManagedIdentity\" or \"SDK\"",
 	},
 	{
 		filename: "azure_bearertoken_basicauth.bad.yml",
@@ -2036,16 +2038,16 @@ func TestExpandExternalLabels(t *testing.T) {
 
 	c, err := LoadFile("testdata/external_labels.good.yml", false, false, log.NewNopLogger())
 	require.NoError(t, err)
-	require.Equal(t, labels.FromStrings("bar", "foo", "baz", "foo${TEST}bar", "foo", "${TEST}", "qux", "foo$${TEST}", "xyz", "foo$$bar"), c.GlobalConfig.ExternalLabels)
+	testutil.RequireEqual(t, labels.FromStrings("bar", "foo", "baz", "foo${TEST}bar", "foo", "${TEST}", "qux", "foo$${TEST}", "xyz", "foo$$bar"), c.GlobalConfig.ExternalLabels)
 
 	c, err = LoadFile("testdata/external_labels.good.yml", false, true, log.NewNopLogger())
 	require.NoError(t, err)
-	require.Equal(t, labels.FromStrings("bar", "foo", "baz", "foobar", "foo", "", "qux", "foo${TEST}", "xyz", "foo$bar"), c.GlobalConfig.ExternalLabels)
+	testutil.RequireEqual(t, labels.FromStrings("bar", "foo", "baz", "foobar", "foo", "", "qux", "foo${TEST}", "xyz", "foo$bar"), c.GlobalConfig.ExternalLabels)
 
 	os.Setenv("TEST", "TestValue")
 	c, err = LoadFile("testdata/external_labels.good.yml", false, true, log.NewNopLogger())
 	require.NoError(t, err)
-	require.Equal(t, labels.FromStrings("bar", "foo", "baz", "fooTestValuebar", "foo", "TestValue", "qux", "foo${TEST}", "xyz", "foo$bar"), c.GlobalConfig.ExternalLabels)
+	testutil.RequireEqual(t, labels.FromStrings("bar", "foo", "baz", "fooTestValuebar", "foo", "TestValue", "qux", "foo${TEST}", "xyz", "foo$bar"), c.GlobalConfig.ExternalLabels)
 }
 
 func TestAgentMode(t *testing.T) {

config/testdata/conf.good.yml

@@ -221,6 +221,7 @@ scrape_configs:
 
     kuma_sd_configs:
       - server: http://kuma-control-plane.kuma-system.svc:5676
+        client_id: main-prometheus
 
   - job_name: service-marathon
     marathon_sd_configs:

config/testdata/roundtrip.good.yml

@@ -108,6 +108,7 @@ scrape_configs:
 
     kuma_sd_configs:
       - server: http://kuma-control-plane.kuma-system.svc:5676
+        client_id: main-prometheus
 
     marathon_sd_configs:
       - servers:

consoles/node-cpu.html

@@ -47,7 +47,7 @@
   <script>
   new PromConsole.Graph({
     node: document.querySelector("#cpuGraph"),
-    expr: "sum by (mode)(irate(node_cpu_seconds_total{job='node',instance='{{ .Params.instance }}',mode!='idle'}[5m]))",
+    expr: "sum by (mode)(irate(node_cpu_seconds_total{job='node',instance='{{ .Params.instance }}',mode!='idle',mode!='iowait',mode!='steal'}[5m]))",
     renderer: 'area',
     max: {{ with printf "count(count by (cpu)(node_cpu_seconds_total{job='node',instance='%s'}))" .Params.instance | query }}{{ . | first | value }}{{ else}}undefined{{end}},
     yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,

consoles/node-overview.html

@@ -69,7 +69,7 @@
   <script>
   new PromConsole.Graph({
     node: document.querySelector("#cpuGraph"),
-    expr: "sum by (mode)(irate(node_cpu_seconds_total{job='node',instance='{{ .Params.instance }}',mode!='idle'}[5m]))",
+    expr: "sum by (mode)(irate(node_cpu_seconds_total{job='node',instance='{{ .Params.instance }}',mode!='idle',mode!='iowait',mode!='steal'}[5m]))",
     renderer: 'area',
     max: {{ with printf "count(count by (cpu)(node_cpu_seconds_total{job='node',instance='%s'}))" .Params.instance | query }}{{ . | first | value }}{{ else}}undefined{{end}},
     yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,

consoles/node.html

@@ -21,7 +21,7 @@
 <tr>
   <td><a href="node-overview.html?instance={{ .Labels.instance }}">{{ reReplaceAll "(.*?://)([^:/]+?)(:\\d+)?/.*" "$2" .Labels.instance }}</a></td>
   <td{{ if eq (. | value) 1.0 }}>Yes{{ else }} class="alert-danger">No{{ end }}</td>
-  <td>{{ template "prom_query_drilldown" (args (printf "100 * (1 - avg by(instance)(irate(node_cpu_seconds_total{job='node',mode='idle',instance='%s'}[5m])))" .Labels.instance) "%" "printf.1f") }}</td>
+  <td>{{ template "prom_query_drilldown" (args (printf "100 * (1 - avg by(instance) (sum without(mode) (irate(node_cpu_seconds_total{job='node',mode=~'idle|iowait|steal',instance='%s'}[5m]))))" .Labels.instance) "%" "printf.1f") }}</td>
   <td>{{ template "prom_query_drilldown" (args (printf "node_memory_MemFree_bytes{job='node',instance='%s'} + node_memory_Cached_bytes{job='node',instance='%s'} + node_memory_Buffers_bytes{job='node',instance='%s'}" .Labels.instance .Labels.instance .Labels.instance) "B" "humanize1024") }}</td>
 </tr>
 {{ else }}

discovery/aws/ec2.go

@@ -18,6 +18,7 @@ import (
 	"errors"
 	"fmt"
 	"net"
+	"strconv"
 	"strings"
 	"time"
 
@@ -97,12 +98,19 @@ type EC2SDConfig struct {
 	HTTPClientConfig config.HTTPClientConfig `yaml:",inline"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*EC2SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &ec2Metrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the EC2 Config.
 func (*EC2SDConfig) Name() string { return "ec2" }
 
 // NewDiscoverer returns a Discoverer for the EC2 Config.
 func (c *EC2SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewEC2Discovery(c, opts.Logger, opts.Registerer), nil
+	return NewEC2Discovery(c, opts.Logger, opts.Metrics)
 }
 
 // UnmarshalYAML implements the yaml.Unmarshaler interface for the EC2 Config.
@@ -148,7 +156,12 @@ type EC2Discovery struct {
 }
 
 // NewEC2Discovery returns a new EC2Discovery which periodically refreshes its targets.
-func NewEC2Discovery(conf *EC2SDConfig, logger log.Logger, reg prometheus.Registerer) *EC2Discovery {
+func NewEC2Discovery(conf *EC2SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*EC2Discovery, error) {
+	m, ok := metrics.(*ec2Metrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -158,14 +171,14 @@ func NewEC2Discovery(conf *EC2SDConfig, logger log.Logger, reg prometheus.Regist
 	}
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "ec2",
-			Interval: time.Duration(d.cfg.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "ec2",
+			Interval:            time.Duration(d.cfg.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
-	return d
+	return d, nil
 }
 
 func (d *EC2Discovery) ec2Client(context.Context) (*ec2.EC2, error) {
@@ -267,7 +280,7 @@ func (d *EC2Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error
 				if inst.PrivateDnsName != nil {
 					labels[ec2LabelPrivateDNS] = model.LabelValue(*inst.PrivateDnsName)
 				}
-				addr := net.JoinHostPort(*inst.PrivateIpAddress, fmt.Sprintf("%d", d.cfg.Port))
+				addr := net.JoinHostPort(*inst.PrivateIpAddress, strconv.Itoa(d.cfg.Port))
 				labels[model.AddressLabel] = model.LabelValue(addr)
 
 				if inst.Platform != nil {

discovery/aws/lightsail.go

@@ -18,6 +18,7 @@ import (
 	"errors"
 	"fmt"
 	"net"
+	"strconv"
 	"strings"
 	"time"
 
@@ -80,12 +81,19 @@ type LightsailSDConfig struct {
 	HTTPClientConfig config.HTTPClientConfig `yaml:",inline"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*LightsailSDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &lightsailMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the Lightsail Config.
 func (*LightsailSDConfig) Name() string { return "lightsail" }
 
 // NewDiscoverer returns a Discoverer for the Lightsail Config.
 func (c *LightsailSDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewLightsailDiscovery(c, opts.Logger, opts.Registerer), nil
+	return NewLightsailDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 // UnmarshalYAML implements the yaml.Unmarshaler interface for the Lightsail Config.
@@ -122,23 +130,29 @@ type LightsailDiscovery struct {
 }
 
 // NewLightsailDiscovery returns a new LightsailDiscovery which periodically refreshes its targets.
-func NewLightsailDiscovery(conf *LightsailSDConfig, logger log.Logger, reg prometheus.Registerer) *LightsailDiscovery {
+func NewLightsailDiscovery(conf *LightsailSDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*LightsailDiscovery, error) {
+	m, ok := metrics.(*lightsailMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
+
 	d := &LightsailDiscovery{
 		cfg: conf,
 	}
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "lightsail",
-			Interval: time.Duration(d.cfg.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "lightsail",
+			Interval:            time.Duration(d.cfg.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
-	return d
+	return d, nil
 }
 
 func (d *LightsailDiscovery) lightsailClient() (*lightsail.Lightsail, error) {
@@ -216,7 +230,7 @@ func (d *LightsailDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group,
 			lightsailLabelRegion:              model.LabelValue(d.cfg.Region),
 		}
 
-		addr := net.JoinHostPort(*inst.PrivateIpAddress, fmt.Sprintf("%d", d.cfg.Port))
+		addr := net.JoinHostPort(*inst.PrivateIpAddress, strconv.Itoa(d.cfg.Port))
 		labels[model.AddressLabel] = model.LabelValue(addr)
 
 		if inst.PublicIpAddress != nil {

discovery/aws/metrics_ec2.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package aws
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+type ec2Metrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+var _ discovery.DiscovererMetrics = (*ec2Metrics)(nil)
+
+// Register implements discovery.DiscovererMetrics.
+func (m *ec2Metrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *ec2Metrics) Unregister() {}

discovery/aws/metrics_lightsail.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package aws
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+type lightsailMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+var _ discovery.DiscovererMetrics = (*lightsailMetrics)(nil)
+
+// Register implements discovery.DiscovererMetrics.
+func (m *lightsailMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *lightsailMetrics) Unregister() {}

discovery/azure/azure.go

@@ -20,6 +20,7 @@ import (
 	"math/rand"
 	"net"
 	"net/http"
+	"strconv"
 	"strings"
 	"sync"
 	"time"
@@ -30,8 +31,8 @@ import (
 	"github.com/Azure/azure-sdk-for-go/sdk/azcore/policy"
 	"github.com/Azure/azure-sdk-for-go/sdk/azcore/to"
 	"github.com/Azure/azure-sdk-for-go/sdk/azidentity"
-	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/compute/armcompute/v4"
-	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/network/armnetwork/v2"
+	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/compute/armcompute/v5"
+	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/network/armnetwork/v4"
 	cache "github.com/Code-Hex/go-generics-cache"
 	"github.com/Code-Hex/go-generics-cache/policy/lru"
 	"github.com/go-kit/log"
@@ -65,6 +66,7 @@ const (
 	azureLabelMachineSize          = azureLabel + "machine_size"
 
 	authMethodOAuth           = "OAuth"
+	authMethodSDK             = "SDK"
 	authMethodManagedIdentity = "ManagedIdentity"
 )
 
@@ -120,12 +122,17 @@ type SDConfig struct {
 	HTTPClientConfig config_util.HTTPClientConfig `yaml:",inline"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "azure" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 func validateAuthParam(param, name string) error {
@@ -159,8 +166,8 @@ func (c *SDConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 		}
 	}
 
-	if c.AuthenticationMethod != authMethodOAuth && c.AuthenticationMethod != authMethodManagedIdentity {
-		return fmt.Errorf("unknown authentication_type %q. Supported types are %q or %q", c.AuthenticationMethod, authMethodOAuth, authMethodManagedIdentity)
+	if c.AuthenticationMethod != authMethodOAuth && c.AuthenticationMethod != authMethodManagedIdentity && c.AuthenticationMethod != authMethodSDK {
+		return fmt.Errorf("unknown authentication_type %q. Supported types are %q, %q or %q", c.AuthenticationMethod, authMethodOAuth, authMethodManagedIdentity, authMethodSDK)
 	}
 
 	return c.HTTPClientConfig.Validate()
@@ -168,51 +175,53 @@ func (c *SDConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 
 type Discovery struct {
 	*refresh.Discovery
-	logger        log.Logger
-	cfg           *SDConfig
-	port          int
-	cache         *cache.Cache[string, *armnetwork.Interface]
-	failuresCount prometheus.Counter
-	cacheHitCount prometheus.Counter
+	logger  log.Logger
+	cfg     *SDConfig
+	port    int
+	cache   *cache.Cache[string, *armnetwork.Interface]
+	metrics *azureMetrics
 }
 
 // NewDiscovery returns a new AzureDiscovery which periodically refreshes its targets.
-func NewDiscovery(cfg *SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(cfg *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*azureMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
 	l := cache.New(cache.AsLRU[string, *armnetwork.Interface](lru.WithCapacity(5000)))
 	d := &Discovery{
-		cfg:    cfg,
-		port:   cfg.Port,
-		logger: logger,
-		cache:  l,
-		failuresCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Name: "prometheus_sd_azure_failures_total",
-				Help: "Number of Azure service discovery refresh failures.",
-			}),
-		cacheHitCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Name: "prometheus_sd_azure_cache_hit_total",
-				Help: "Number of cache hit during refresh.",
-			}),
+		cfg:     cfg,
+		port:    cfg.Port,
+		logger:  logger,
+		cache:   l,
+		metrics: m,
 	}
 
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "azure",
-			Interval: time.Duration(cfg.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
-			Metrics:  []prometheus.Collector{d.failuresCount, d.cacheHitCount},
+			Logger:              logger,
+			Mech:                "azure",
+			Interval:            time.Duration(cfg.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 
 	return d, nil
 }
 
+type client interface {
+	getVMs(ctx context.Context, resourceGroup string) ([]virtualMachine, error)
+	getScaleSets(ctx context.Context, resourceGroup string) ([]armcompute.VirtualMachineScaleSet, error)
+	getScaleSetVMs(ctx context.Context, scaleSet armcompute.VirtualMachineScaleSet) ([]virtualMachine, error)
+	getVMNetworkInterfaceByID(ctx context.Context, networkInterfaceID string) (*armnetwork.Interface, error)
+	getVMScaleSetVMNetworkInterfaceByID(ctx context.Context, networkInterfaceID, scaleSetName, instanceID string) (*armnetwork.Interface, error)
+}
+
 // azureClient represents multiple Azure Resource Manager providers.
 type azureClient struct {
 	nic    *armnetwork.InterfacesClient
@@ -222,14 +231,17 @@ type azureClient struct {
 	logger log.Logger
 }
 
+var _ client = &azureClient{}
+
 // createAzureClient is a helper function for creating an Azure compute client to ARM.
-func createAzureClient(cfg SDConfig) (azureClient, error) {
+func createAzureClient(cfg SDConfig, logger log.Logger) (client, error) {
 	cloudConfiguration, err := CloudConfigurationFromName(cfg.Environment)
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
 	var c azureClient
+	c.logger = logger
 
 	telemetry := policy.TelemetryOptions{
 		ApplicationID: userAgent,
@@ -240,12 +252,12 @@ func createAzureClient(cfg SDConfig) (azureClient, error) {
 		Telemetry: telemetry,
 	})
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
 	client, err := config_util.NewClientFromConfig(cfg.HTTPClientConfig, "azure_sd")
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 	options := &arm.ClientOptions{
 		ClientOptions: policy.ClientOptions{
@@ -257,25 +269,25 @@ func createAzureClient(cfg SDConfig) (azureClient, error) {
 
 	c.vm, err = armcompute.NewVirtualMachinesClient(cfg.SubscriptionID, credential, options)
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
 	c.nic, err = armnetwork.NewInterfacesClient(cfg.SubscriptionID, credential, options)
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
 	c.vmss, err = armcompute.NewVirtualMachineScaleSetsClient(cfg.SubscriptionID, credential, options)
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
 	c.vmssvm, err = armcompute.NewVirtualMachineScaleSetVMsClient(cfg.SubscriptionID, credential, options)
 	if err != nil {
-		return azureClient{}, err
+		return &azureClient{}, err
 	}
 
-	return c, nil
+	return &c, nil
 }
 
 func newCredential(cfg SDConfig, policyClientOptions policy.ClientOptions) (azcore.TokenCredential, error) {
@@ -295,6 +307,16 @@ func newCredential(cfg SDConfig, policyClientOptions policy.ClientOptions) (azco
 			return nil, err
 		}
 		credential = azcore.TokenCredential(secretCredential)
+	case authMethodSDK:
+		options := &azidentity.DefaultAzureCredentialOptions{ClientOptions: policyClientOptions}
+		if len(cfg.TenantID) != 0 {
+			options.TenantID = cfg.TenantID
+		}
+		sdkCredential, err := azidentity.NewDefaultAzureCredential(options)
+		if err != nil {
+			return nil, err
+		}
+		credential = azcore.TokenCredential(sdkCredential)
 	}
 	return credential, nil
 }
@@ -331,16 +353,15 @@ func newAzureResourceFromID(id string, logger log.Logger) (*arm.ResourceID, erro
 func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	defer level.Debug(d.logger).Log("msg", "Azure discovery completed")
 
-	client, err := createAzureClient(*d.cfg)
+	client, err := createAzureClient(*d.cfg, d.logger)
 	if err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, fmt.Errorf("could not create Azure client: %w", err)
 	}
-	client.logger = d.logger
 
 	machines, err := client.getVMs(ctx, d.cfg.ResourceGroup)
 	if err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, fmt.Errorf("could not get virtual machines: %w", err)
 	}
 
@@ -349,14 +370,14 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	// Load the vms managed by scale sets.
 	scaleSets, err := client.getScaleSets(ctx, d.cfg.ResourceGroup)
 	if err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, fmt.Errorf("could not get virtual machine scale sets: %w", err)
 	}
 
 	for _, scaleSet := range scaleSets {
 		scaleSetVms, err := client.getScaleSetVMs(ctx, scaleSet)
 		if err != nil {
-			d.failuresCount.Inc()
+			d.metrics.failuresCount.Inc()
 			return nil, fmt.Errorf("could not get virtual machine scale set vms: %w", err)
 		}
 		machines = append(machines, scaleSetVms...)
@@ -375,102 +396,8 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	for _, vm := range machines {
 		go func(vm virtualMachine) {
 			defer wg.Done()
-			r, err := newAzureResourceFromID(vm.ID, d.logger)
-			if err != nil {
-				ch <- target{labelSet: nil, err: err}
-				return
-			}
-
-			labels := model.LabelSet{
-				azureLabelSubscriptionID:       model.LabelValue(d.cfg.SubscriptionID),
-				azureLabelTenantID:             model.LabelValue(d.cfg.TenantID),
-				azureLabelMachineID:            model.LabelValue(vm.ID),
-				azureLabelMachineName:          model.LabelValue(vm.Name),
-				azureLabelMachineComputerName:  model.LabelValue(vm.ComputerName),
-				azureLabelMachineOSType:        model.LabelValue(vm.OsType),
-				azureLabelMachineLocation:      model.LabelValue(vm.Location),
-				azureLabelMachineResourceGroup: model.LabelValue(r.ResourceGroupName),
-				azureLabelMachineSize:          model.LabelValue(vm.Size),
-			}
-
-			if vm.ScaleSet != "" {
-				labels[azureLabelMachineScaleSet] = model.LabelValue(vm.ScaleSet)
-			}
-
-			for k, v := range vm.Tags {
-				name := strutil.SanitizeLabelName(k)
-				labels[azureLabelMachineTag+model.LabelName(name)] = model.LabelValue(*v)
-			}
-
-			// Get the IP address information via separate call to the network provider.
-			for _, nicID := range vm.NetworkInterfaces {
-				var networkInterface *armnetwork.Interface
-				if v, ok := d.getFromCache(nicID); ok {
-					networkInterface = v
-					d.cacheHitCount.Add(1)
-				} else {
-					if vm.ScaleSet == "" {
-						networkInterface, err = client.getVMNetworkInterfaceByID(ctx, nicID)
-						if err != nil {
-							if errors.Is(err, errorNotFound) {
-								level.Warn(d.logger).Log("msg", "Network interface does not exist", "name", nicID, "err", err)
-							} else {
-								ch <- target{labelSet: nil, err: err}
-							}
-							// Get out of this routine because we cannot continue without a network interface.
-							return
-						}
-						d.addToCache(nicID, networkInterface)
-					} else {
-						networkInterface, err = client.getVMScaleSetVMNetworkInterfaceByID(ctx, nicID, vm.ScaleSet, vm.InstanceID)
-						if err != nil {
-							if errors.Is(err, errorNotFound) {
-								level.Warn(d.logger).Log("msg", "Network interface does not exist", "name", nicID, "err", err)
-							} else {
-								ch <- target{labelSet: nil, err: err}
-							}
-							// Get out of this routine because we cannot continue without a network interface.
-							return
-						}
-						d.addToCache(nicID, networkInterface)
-					}
-				}
-
-				if networkInterface.Properties == nil {
-					continue
-				}
-
-				// Unfortunately Azure does not return information on whether a VM is deallocated.
-				// This information is available via another API call however the Go SDK does not
-				// yet support this. On deallocated machines, this value happens to be nil so it
-				// is a cheap and easy way to determine if a machine is allocated or not.
-				if networkInterface.Properties.Primary == nil {
-					level.Debug(d.logger).Log("msg", "Skipping deallocated virtual machine", "machine", vm.Name)
-					return
-				}
-
-				if *networkInterface.Properties.Primary {
-					for _, ip := range networkInterface.Properties.IPConfigurations {
-						// IPAddress is a field defined in PublicIPAddressPropertiesFormat,
-						// therefore we need to validate that both are not nil.
-						if ip.Properties != nil && ip.Properties.PublicIPAddress != nil && ip.Properties.PublicIPAddress.Properties != nil && ip.Properties.PublicIPAddress.Properties.IPAddress != nil {
-							labels[azureLabelMachinePublicIP] = model.LabelValue(*ip.Properties.PublicIPAddress.Properties.IPAddress)
-						}
-						if ip.Properties != nil && ip.Properties.PrivateIPAddress != nil {
-							labels[azureLabelMachinePrivateIP] = model.LabelValue(*ip.Properties.PrivateIPAddress)
-							address := net.JoinHostPort(*ip.Properties.PrivateIPAddress, fmt.Sprintf("%d", d.port))
-							labels[model.AddressLabel] = model.LabelValue(address)
-							ch <- target{labelSet: labels, err: nil}
-							return
-						}
-						// If we made it here, we don't have a private IP which should be impossible.
-						// Return an empty target and error to ensure an all or nothing situation.
-						err = fmt.Errorf("unable to find a private IP for VM %s", vm.Name)
-						ch <- target{labelSet: nil, err: err}
-						return
-					}
-				}
-			}
+			labelSet, err := d.vmToLabelSet(ctx, client, vm)
+			ch <- target{labelSet: labelSet, err: err}
 		}(vm)
 	}
 
@@ -480,7 +407,7 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	var tg targetgroup.Group
 	for tgt := range ch {
 		if tgt.err != nil {
-			d.failuresCount.Inc()
+			d.metrics.failuresCount.Inc()
 			return nil, fmt.Errorf("unable to complete Azure service discovery: %w", tgt.err)
 		}
 		if tgt.labelSet != nil {
@@ -491,6 +418,95 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	return []*targetgroup.Group{&tg}, nil
 }
 
+func (d *Discovery) vmToLabelSet(ctx context.Context, client client, vm virtualMachine) (model.LabelSet, error) {
+	r, err := newAzureResourceFromID(vm.ID, d.logger)
+	if err != nil {
+		return nil, err
+	}
+
+	labels := model.LabelSet{
+		azureLabelSubscriptionID:       model.LabelValue(d.cfg.SubscriptionID),
+		azureLabelTenantID:             model.LabelValue(d.cfg.TenantID),
+		azureLabelMachineID:            model.LabelValue(vm.ID),
+		azureLabelMachineName:          model.LabelValue(vm.Name),
+		azureLabelMachineComputerName:  model.LabelValue(vm.ComputerName),
+		azureLabelMachineOSType:        model.LabelValue(vm.OsType),
+		azureLabelMachineLocation:      model.LabelValue(vm.Location),
+		azureLabelMachineResourceGroup: model.LabelValue(r.ResourceGroupName),
+		azureLabelMachineSize:          model.LabelValue(vm.Size),
+	}
+
+	if vm.ScaleSet != "" {
+		labels[azureLabelMachineScaleSet] = model.LabelValue(vm.ScaleSet)
+	}
+
+	for k, v := range vm.Tags {
+		name := strutil.SanitizeLabelName(k)
+		labels[azureLabelMachineTag+model.LabelName(name)] = model.LabelValue(*v)
+	}
+
+	// Get the IP address information via separate call to the network provider.
+	for _, nicID := range vm.NetworkInterfaces {
+		var networkInterface *armnetwork.Interface
+		if v, ok := d.getFromCache(nicID); ok {
+			networkInterface = v
+			d.metrics.cacheHitCount.Add(1)
+		} else {
+			if vm.ScaleSet == "" {
+				networkInterface, err = client.getVMNetworkInterfaceByID(ctx, nicID)
+			} else {
+				networkInterface, err = client.getVMScaleSetVMNetworkInterfaceByID(ctx, nicID, vm.ScaleSet, vm.InstanceID)
+			}
+			if err != nil {
+				if errors.Is(err, errorNotFound) {
+					level.Warn(d.logger).Log("msg", "Network interface does not exist", "name", nicID, "err", err)
+				} else {
+					return nil, err
+				}
+				// Get out of this routine because we cannot continue without a network interface.
+				return nil, nil
+			}
+
+			// Continue processing with the network interface
+			d.addToCache(nicID, networkInterface)
+		}
+
+		if networkInterface.Properties == nil {
+			continue
+		}
+
+		// Unfortunately Azure does not return information on whether a VM is deallocated.
+		// This information is available via another API call however the Go SDK does not
+		// yet support this. On deallocated machines, this value happens to be nil so it
+		// is a cheap and easy way to determine if a machine is allocated or not.
+		if networkInterface.Properties.Primary == nil {
+			level.Debug(d.logger).Log("msg", "Skipping deallocated virtual machine", "machine", vm.Name)
+			return nil, nil
+		}
+
+		if *networkInterface.Properties.Primary {
+			for _, ip := range networkInterface.Properties.IPConfigurations {
+				// IPAddress is a field defined in PublicIPAddressPropertiesFormat,
+				// therefore we need to validate that both are not nil.
+				if ip.Properties != nil && ip.Properties.PublicIPAddress != nil && ip.Properties.PublicIPAddress.Properties != nil && ip.Properties.PublicIPAddress.Properties.IPAddress != nil {
+					labels[azureLabelMachinePublicIP] = model.LabelValue(*ip.Properties.PublicIPAddress.Properties.IPAddress)
+				}
+				if ip.Properties != nil && ip.Properties.PrivateIPAddress != nil {
+					labels[azureLabelMachinePrivateIP] = model.LabelValue(*ip.Properties.PrivateIPAddress)
+					address := net.JoinHostPort(*ip.Properties.PrivateIPAddress, strconv.Itoa(d.port))
+					labels[model.AddressLabel] = model.LabelValue(address)
+					return labels, nil
+				}
+				// If we made it here, we don't have a private IP which should be impossible.
+				// Return an empty target and error to ensure an all or nothing situation.
+				return nil, fmt.Errorf("unable to find a private IP for VM %s", vm.Name)
+			}
+		}
+	}
+	// TODO: Should we say something at this point?
+	return nil, nil
+}
+
 func (client *azureClient) getVMs(ctx context.Context, resourceGroup string) ([]virtualMachine, error) {
 	var vms []virtualMachine
 	if len(resourceGroup) == 0 {
@@ -570,7 +586,7 @@ func (client *azureClient) getScaleSetVMs(ctx context.Context, scaleSet armcompu
 }
 
 func mapFromVM(vm armcompute.VirtualMachine) virtualMachine {
-	osType := string(*vm.Properties.StorageProfile.OSDisk.OSType)
+	var osType string
 	tags := map[string]*string{}
 	networkInterfaces := []string{}
 	var computerName string
@@ -581,6 +597,12 @@ func mapFromVM(vm armcompute.VirtualMachine) virtualMachine {
 	}
 
 	if vm.Properties != nil {
+		if vm.Properties.StorageProfile != nil &&
+			vm.Properties.StorageProfile.OSDisk != nil &&
+			vm.Properties.StorageProfile.OSDisk.OSType != nil {
+			osType = string(*vm.Properties.StorageProfile.OSDisk.OSType)
+		}
+
 		if vm.Properties.NetworkProfile != nil {
 			for _, vmNIC := range vm.Properties.NetworkProfile.NetworkInterfaces {
 				networkInterfaces = append(networkInterfaces, *vmNIC.ID)
@@ -609,7 +631,7 @@ func mapFromVM(vm armcompute.VirtualMachine) virtualMachine {
 }
 
 func mapFromVMScaleSetVM(vm armcompute.VirtualMachineScaleSetVM, scaleSetName string) virtualMachine {
-	osType := string(*vm.Properties.StorageProfile.OSDisk.OSType)
+	var osType string
 	tags := map[string]*string{}
 	networkInterfaces := []string{}
 	var computerName string
@@ -620,6 +642,12 @@ func mapFromVMScaleSetVM(vm armcompute.VirtualMachineScaleSetVM, scaleSetName st
 	}
 
 	if vm.Properties != nil {
+		if vm.Properties.StorageProfile != nil &&
+			vm.Properties.StorageProfile.OSDisk != nil &&
+			vm.Properties.StorageProfile.OSDisk.OSType != nil {
+			osType = string(*vm.Properties.StorageProfile.OSDisk.OSType)
+		}
+
 		if vm.Properties.NetworkProfile != nil {
 			for _, vmNIC := range vm.Properties.NetworkProfile.NetworkInterfaces {
 				networkInterfaces = append(networkInterfaces, *vmNIC.ID)

discovery/azure/azure_test.go

@@ -14,16 +14,24 @@
 package azure
 
 import (
+	"context"
+	"fmt"
 	"testing"
 
 	"github.com/Azure/azure-sdk-for-go/sdk/azcore/arm"
-	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/compute/armcompute/v4"
+	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/compute/armcompute/v5"
+	"github.com/Azure/azure-sdk-for-go/sdk/resourcemanager/network/armnetwork/v4"
+	cache "github.com/Code-Hex/go-generics-cache"
+	"github.com/Code-Hex/go-generics-cache/policy/lru"
+	"github.com/go-kit/log"
 	"github.com/stretchr/testify/require"
 	"go.uber.org/goleak"
 )
 
 func TestMain(m *testing.M) {
-	goleak.VerifyTestMain(m)
+	goleak.VerifyTestMain(m,
+		goleak.IgnoreTopFunction("github.com/Code-Hex/go-generics-cache.(*janitor).run.func1"),
+	)
 }
 
 func TestMapFromVMWithEmptyTags(t *testing.T) {
@@ -79,6 +87,140 @@ func TestMapFromVMWithEmptyTags(t *testing.T) {
 	require.Equal(t, expectedVM, actualVM)
 }
 
+func TestVMToLabelSet(t *testing.T) {
+	id := "/subscriptions/00000000-0000-0000-0000-000000000000/test"
+	name := "name"
+	size := "size"
+	vmSize := armcompute.VirtualMachineSizeTypes(size)
+	osType := armcompute.OperatingSystemTypesLinux
+	vmType := "type"
+	location := "westeurope"
+	computerName := "computer_name"
+	networkID := "/subscriptions/00000000-0000-0000-0000-000000000000/network1"
+	ipAddress := "10.20.30.40"
+	primary := true
+	networkProfile := armcompute.NetworkProfile{
+		NetworkInterfaces: []*armcompute.NetworkInterfaceReference{
+			{
+				ID:         &networkID,
+				Properties: &armcompute.NetworkInterfaceReferenceProperties{Primary: &primary},
+			},
+		},
+	}
+	properties := &armcompute.VirtualMachineProperties{
+		OSProfile: &armcompute.OSProfile{
+			ComputerName: &computerName,
+		},
+		StorageProfile: &armcompute.StorageProfile{
+			OSDisk: &armcompute.OSDisk{
+				OSType: &osType,
+			},
+		},
+		NetworkProfile: &networkProfile,
+		HardwareProfile: &armcompute.HardwareProfile{
+			VMSize: &vmSize,
+		},
+	}
+
+	testVM := armcompute.VirtualMachine{
+		ID:         &id,
+		Name:       &name,
+		Type:       &vmType,
+		Location:   &location,
+		Tags:       nil,
+		Properties: properties,
+	}
+
+	expectedVM := virtualMachine{
+		ID:                id,
+		Name:              name,
+		ComputerName:      computerName,
+		Type:              vmType,
+		Location:          location,
+		OsType:            "Linux",
+		Tags:              map[string]*string{},
+		NetworkInterfaces: []string{networkID},
+		Size:              size,
+	}
+
+	actualVM := mapFromVM(testVM)
+
+	require.Equal(t, expectedVM, actualVM)
+
+	cfg := DefaultSDConfig
+	d := &Discovery{
+		cfg:    &cfg,
+		logger: log.NewNopLogger(),
+		cache:  cache.New(cache.AsLRU[string, *armnetwork.Interface](lru.WithCapacity(5))),
+	}
+	network := armnetwork.Interface{
+		Name: &networkID,
+		Properties: &armnetwork.InterfacePropertiesFormat{
+			Primary: &primary,
+			IPConfigurations: []*armnetwork.InterfaceIPConfiguration{
+				{Properties: &armnetwork.InterfaceIPConfigurationPropertiesFormat{
+					PrivateIPAddress: &ipAddress,
+				}},
+			},
+		},
+	}
+	client := &mockAzureClient{
+		networkInterface: &network,
+	}
+	labelSet, err := d.vmToLabelSet(context.Background(), client, actualVM)
+	require.NoError(t, err)
+	require.Len(t, labelSet, 11)
+}
+
+func TestMapFromVMWithEmptyOSType(t *testing.T) {
+	id := "test"
+	name := "name"
+	size := "size"
+	vmSize := armcompute.VirtualMachineSizeTypes(size)
+	vmType := "type"
+	location := "westeurope"
+	computerName := "computer_name"
+	networkProfile := armcompute.NetworkProfile{
+		NetworkInterfaces: []*armcompute.NetworkInterfaceReference{},
+	}
+	properties := &armcompute.VirtualMachineProperties{
+		OSProfile: &armcompute.OSProfile{
+			ComputerName: &computerName,
+		},
+		StorageProfile: &armcompute.StorageProfile{
+			OSDisk: &armcompute.OSDisk{},
+		},
+		NetworkProfile: &networkProfile,
+		HardwareProfile: &armcompute.HardwareProfile{
+			VMSize: &vmSize,
+		},
+	}
+
+	testVM := armcompute.VirtualMachine{
+		ID:         &id,
+		Name:       &name,
+		Type:       &vmType,
+		Location:   &location,
+		Tags:       nil,
+		Properties: properties,
+	}
+
+	expectedVM := virtualMachine{
+		ID:                id,
+		Name:              name,
+		ComputerName:      computerName,
+		Type:              vmType,
+		Location:          location,
+		Tags:              map[string]*string{},
+		NetworkInterfaces: []string{},
+		Size:              size,
+	}
+
+	actualVM := mapFromVM(testVM)
+
+	require.Equal(t, expectedVM, actualVM)
+}
+
 func TestMapFromVMWithTags(t *testing.T) {
 	id := "test"
 	name := "name"
@@ -193,6 +335,58 @@ func TestMapFromVMScaleSetVMWithEmptyTags(t *testing.T) {
 	require.Equal(t, expectedVM, actualVM)
 }
 
+func TestMapFromVMScaleSetVMWithEmptyOSType(t *testing.T) {
+	id := "test"
+	name := "name"
+	size := "size"
+	vmSize := armcompute.VirtualMachineSizeTypes(size)
+	vmType := "type"
+	instanceID := "123"
+	location := "westeurope"
+	computerName := "computer_name"
+	networkProfile := armcompute.NetworkProfile{
+		NetworkInterfaces: []*armcompute.NetworkInterfaceReference{},
+	}
+	properties := &armcompute.VirtualMachineScaleSetVMProperties{
+		OSProfile: &armcompute.OSProfile{
+			ComputerName: &computerName,
+		},
+		StorageProfile: &armcompute.StorageProfile{},
+		NetworkProfile: &networkProfile,
+		HardwareProfile: &armcompute.HardwareProfile{
+			VMSize: &vmSize,
+		},
+	}
+
+	testVM := armcompute.VirtualMachineScaleSetVM{
+		ID:         &id,
+		Name:       &name,
+		Type:       &vmType,
+		InstanceID: &instanceID,
+		Location:   &location,
+		Tags:       nil,
+		Properties: properties,
+	}
+
+	scaleSet := "testSet"
+	expectedVM := virtualMachine{
+		ID:                id,
+		Name:              name,
+		ComputerName:      computerName,
+		Type:              vmType,
+		Location:          location,
+		Tags:              map[string]*string{},
+		NetworkInterfaces: []string{},
+		ScaleSet:          scaleSet,
+		InstanceID:        instanceID,
+		Size:              size,
+	}
+
+	actualVM := mapFromVMScaleSetVM(testVM, scaleSet)
+
+	require.Equal(t, expectedVM, actualVM)
+}
+
 func TestMapFromVMScaleSetVMWithTags(t *testing.T) {
 	id := "test"
 	name := "name"
@@ -280,3 +474,35 @@ func TestNewAzureResourceFromID(t *testing.T) {
 		require.Equal(t, tc.expected.ResourceGroupName, actual.ResourceGroupName)
 	}
 }
+
+type mockAzureClient struct {
+	networkInterface *armnetwork.Interface
+}
+
+var _ client = &mockAzureClient{}
+
+func (*mockAzureClient) getVMs(ctx context.Context, resourceGroup string) ([]virtualMachine, error) {
+	return nil, nil
+}
+
+func (*mockAzureClient) getScaleSets(ctx context.Context, resourceGroup string) ([]armcompute.VirtualMachineScaleSet, error) {
+	return nil, nil
+}
+
+func (*mockAzureClient) getScaleSetVMs(ctx context.Context, scaleSet armcompute.VirtualMachineScaleSet) ([]virtualMachine, error) {
+	return nil, nil
+}
+
+func (m *mockAzureClient) getVMNetworkInterfaceByID(ctx context.Context, networkInterfaceID string) (*armnetwork.Interface, error) {
+	if networkInterfaceID == "" {
+		return nil, fmt.Errorf("parameter networkInterfaceID cannot be empty")
+	}
+	return m.networkInterface, nil
+}
+
+func (m *mockAzureClient) getVMScaleSetVMNetworkInterfaceByID(ctx context.Context, networkInterfaceID, scaleSetName, instanceID string) (*armnetwork.Interface, error) {
+	if scaleSetName == "" {
+		return nil, fmt.Errorf("parameter virtualMachineScaleSetName cannot be empty")
+	}
+	return m.networkInterface, nil
+}

discovery/azure/metrics.go

@@ -0,0 +1,64 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package azure
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*azureMetrics)(nil)
+
+type azureMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+
+	failuresCount prometheus.Counter
+	cacheHitCount prometheus.Counter
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	m := &azureMetrics{
+		refreshMetrics: rmi,
+		failuresCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Name: "prometheus_sd_azure_failures_total",
+				Help: "Number of Azure service discovery refresh failures.",
+			}),
+		cacheHitCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Name: "prometheus_sd_azure_cache_hit_total",
+				Help: "Number of cache hit during refresh.",
+			}),
+	}
+
+	m.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		m.failuresCount,
+		m.cacheHitCount,
+	})
+
+	return m
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *azureMetrics) Register() error {
+	return m.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *azureMetrics) Unregister() {
+	m.metricRegisterer.UnregisterMetrics()
+}

discovery/consul/consul.go

@@ -119,12 +119,17 @@ type SDConfig struct {
 	HTTPClientConfig config.HTTPClientConfig `yaml:",inline"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "consul" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -161,27 +166,28 @@ func (c *SDConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 // Discovery retrieves target information from a Consul server
 // and updates them via watches.
 type Discovery struct {
-	client              *consul.Client
-	clientDatacenter    string
-	clientNamespace     string
-	clientPartition     string
-	tagSeparator        string
-	watchedServices     []string // Set of services which will be discovered.
-	watchedTags         []string // Tags used to filter instances of a service.
-	watchedNodeMeta     map[string]string
-	allowStale          bool
-	refreshInterval     time.Duration
-	finalizer           func()
-	logger              log.Logger
-	rpcFailuresCount    prometheus.Counter
-	rpcDuration         *prometheus.SummaryVec
-	servicesRPCDuration prometheus.Observer
-	serviceRPCDuration  prometheus.Observer
-	metricRegisterer    discovery.MetricRegisterer
+	client           *consul.Client
+	clientDatacenter string
+	clientNamespace  string
+	clientPartition  string
+	tagSeparator     string
+	watchedServices  []string // Set of services which will be discovered.
+	watchedTags      []string // Tags used to filter instances of a service.
+	watchedNodeMeta  map[string]string
+	allowStale       bool
+	refreshInterval  time.Duration
+	finalizer        func()
+	logger           log.Logger
+	metrics          *consulMetrics
 }
 
 // NewDiscovery returns a new Discovery for the given config.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*consulMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -219,35 +225,9 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 		clientPartition:  conf.Partition,
 		finalizer:        wrapper.CloseIdleConnections,
 		logger:           logger,
-		rpcFailuresCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Namespace: namespace,
-				Name:      "sd_consul_rpc_failures_total",
-				Help:      "The number of Consul RPC call failures.",
-			}),
-		rpcDuration: prometheus.NewSummaryVec(
-			prometheus.SummaryOpts{
-				Namespace:  namespace,
-				Name:       "sd_consul_rpc_duration_seconds",
-				Help:       "The duration of a Consul RPC call in seconds.",
-				Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
-			},
-			[]string{"endpoint", "call"},
-		),
+		metrics:          m,
 	}
 
-	cd.metricRegisterer = discovery.NewMetricRegisterer(
-		reg,
-		[]prometheus.Collector{
-			cd.rpcFailuresCount,
-			cd.rpcDuration,
-		},
-	)
-
-	// Initialize metric vectors.
-	cd.servicesRPCDuration = cd.rpcDuration.WithLabelValues("catalog", "services")
-	cd.serviceRPCDuration = cd.rpcDuration.WithLabelValues("catalog", "service")
-
 	return cd, nil
 }
 
@@ -303,7 +283,7 @@ func (d *Discovery) getDatacenter() error {
 	info, err := d.client.Agent().Self()
 	if err != nil {
 		level.Error(d.logger).Log("msg", "Error retrieving datacenter name", "err", err)
-		d.rpcFailuresCount.Inc()
+		d.metrics.rpcFailuresCount.Inc()
 		return err
 	}
 
@@ -344,13 +324,6 @@ func (d *Discovery) initialize(ctx context.Context) {
 
 // Run implements the Discoverer interface.
 func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
-	err := d.metricRegisterer.RegisterMetrics()
-	if err != nil {
-		level.Error(d.logger).Log("msg", "Unable to register metrics", "err", err.Error())
-		return
-	}
-	defer d.metricRegisterer.UnregisterMetrics()
-
 	if d.finalizer != nil {
 		defer d.finalizer()
 	}
@@ -399,7 +372,7 @@ func (d *Discovery) watchServices(ctx context.Context, ch chan<- []*targetgroup.
 	t0 := time.Now()
 	srvs, meta, err := catalog.Services(opts.WithContext(ctx))
 	elapsed := time.Since(t0)
-	d.servicesRPCDuration.Observe(elapsed.Seconds())
+	d.metrics.servicesRPCDuration.Observe(elapsed.Seconds())
 
 	// Check the context before in order to exit early.
 	select {
@@ -410,7 +383,7 @@ func (d *Discovery) watchServices(ctx context.Context, ch chan<- []*targetgroup.
 
 	if err != nil {
 		level.Error(d.logger).Log("msg", "Error refreshing service list", "err", err)
-		d.rpcFailuresCount.Inc()
+		d.metrics.rpcFailuresCount.Inc()
 		time.Sleep(retryInterval)
 		return
 	}
@@ -490,8 +463,8 @@ func (d *Discovery) watchService(ctx context.Context, ch chan<- []*targetgroup.G
 		},
 		tagSeparator:       d.tagSeparator,
 		logger:             d.logger,
-		rpcFailuresCount:   d.rpcFailuresCount,
-		serviceRPCDuration: d.serviceRPCDuration,
+		rpcFailuresCount:   d.metrics.rpcFailuresCount,
+		serviceRPCDuration: d.metrics.serviceRPCDuration,
 	}
 
 	go func() {
@@ -566,9 +539,9 @@ func (srv *consulService) watch(ctx context.Context, ch chan<- []*targetgroup.Gr
 		// since the service may be registered remotely through a different node.
 		var addr string
 		if serviceNode.Service.Address != "" {
-			addr = net.JoinHostPort(serviceNode.Service.Address, fmt.Sprintf("%d", serviceNode.Service.Port))
+			addr = net.JoinHostPort(serviceNode.Service.Address, strconv.Itoa(serviceNode.Service.Port))
 		} else {
-			addr = net.JoinHostPort(serviceNode.Node.Address, fmt.Sprintf("%d", serviceNode.Service.Port))
+			addr = net.JoinHostPort(serviceNode.Node.Address, strconv.Itoa(serviceNode.Service.Port))
 		}
 
 		labels := model.LabelSet{

discovery/consul/consul_test.go

@@ -29,6 +29,7 @@ import (
 	"go.uber.org/goleak"
 	"gopkg.in/yaml.v2"
 
+	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
@@ -36,20 +37,30 @@ func TestMain(m *testing.M) {
 	goleak.VerifyTestMain(m)
 }
 
+// TODO: Add ability to unregister metrics?
+func NewTestMetrics(t *testing.T, conf discovery.Config, reg prometheus.Registerer) discovery.DiscovererMetrics {
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	require.NoError(t, refreshMetrics.Register())
+
+	metrics := conf.NewDiscovererMetrics(prometheus.NewRegistry(), refreshMetrics)
+	require.NoError(t, metrics.Register())
+
+	return metrics
+}
+
 func TestConfiguredService(t *testing.T) {
 	conf := &SDConfig{
 		Services: []string{"configuredServiceName"},
 	}
-	consulDiscovery, err := NewDiscovery(conf, nil, prometheus.NewRegistry())
-	if err != nil {
-		t.Errorf("Unexpected error when initializing discovery %v", err)
-	}
-	if !consulDiscovery.shouldWatch("configuredServiceName", []string{""}) {
-		t.Errorf("Expected service %s to be watched", "configuredServiceName")
-	}
-	if consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}) {
-		t.Errorf("Expected service %s to not be watched", "nonConfiguredServiceName")
-	}
+
+	metrics := NewTestMetrics(t, conf, prometheus.NewRegistry())
+
+	consulDiscovery, err := NewDiscovery(conf, nil, metrics)
+	require.NoError(t, err, "when initializing discovery")
+	require.True(t, consulDiscovery.shouldWatch("configuredServiceName", []string{""}),
+		"Expected service %s to be watched", "configuredServiceName")
+	require.False(t, consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}),
+		"Expected service %s to not be watched", "nonConfiguredServiceName")
 }
 
 func TestConfiguredServiceWithTag(t *testing.T) {
@@ -57,22 +68,22 @@ func TestConfiguredServiceWithTag(t *testing.T) {
 		Services:    []string{"configuredServiceName"},
 		ServiceTags: []string{"http"},
 	}
-	consulDiscovery, err := NewDiscovery(conf, nil, prometheus.NewRegistry())
-	if err != nil {
-		t.Errorf("Unexpected error when initializing discovery %v", err)
-	}
-	if consulDiscovery.shouldWatch("configuredServiceName", []string{""}) {
-		t.Errorf("Expected service %s to not be watched without tag", "configuredServiceName")
-	}
-	if !consulDiscovery.shouldWatch("configuredServiceName", []string{"http"}) {
-		t.Errorf("Expected service %s to be watched with tag %s", "configuredServiceName", "http")
-	}
-	if consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}) {
-		t.Errorf("Expected service %s to not be watched without tag", "nonConfiguredServiceName")
-	}
-	if consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{"http"}) {
-		t.Errorf("Expected service %s to not be watched with tag %s", "nonConfiguredServiceName", "http")
-	}
+
+	metrics := NewTestMetrics(t, conf, prometheus.NewRegistry())
+
+	consulDiscovery, err := NewDiscovery(conf, nil, metrics)
+	require.NoError(t, err, "when initializing discovery")
+	require.False(t, consulDiscovery.shouldWatch("configuredServiceName", []string{""}),
+		"Expected service %s to not be watched without tag", "configuredServiceName")
+
+	require.True(t, consulDiscovery.shouldWatch("configuredServiceName", []string{"http"}),
+		"Expected service %s to be watched with tag %s", "configuredServiceName", "http")
+
+	require.False(t, consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}),
+		"Expected service %s to not be watched without tag", "nonConfiguredServiceName")
+
+	require.False(t, consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{"http"}),
+		"Expected service %s to not be watched with tag %s", "nonConfiguredServiceName", "http")
 }
 
 func TestConfiguredServiceWithTags(t *testing.T) {
@@ -152,27 +163,24 @@ func TestConfiguredServiceWithTags(t *testing.T) {
 	}
 
 	for _, tc := range cases {
-		consulDiscovery, err := NewDiscovery(tc.conf, nil, prometheus.NewRegistry())
-		if err != nil {
-			t.Errorf("Unexpected error when initializing discovery %v", err)
-		}
-		ret := consulDiscovery.shouldWatch(tc.serviceName, tc.serviceTags)
-		if ret != tc.shouldWatch {
-			t.Errorf("Expected should watch? %t, got %t. Watched service and tags: %s %+v, input was %s %+v", tc.shouldWatch, ret, tc.conf.Services, tc.conf.ServiceTags, tc.serviceName, tc.serviceTags)
-		}
+		metrics := NewTestMetrics(t, tc.conf, prometheus.NewRegistry())
 
+		consulDiscovery, err := NewDiscovery(tc.conf, nil, metrics)
+		require.NoError(t, err, "when initializing discovery")
+		ret := consulDiscovery.shouldWatch(tc.serviceName, tc.serviceTags)
+		require.Equal(t, tc.shouldWatch, ret, "Watched service and tags: %s %+v, input was %s %+v",
+			tc.conf.Services, tc.conf.ServiceTags, tc.serviceName, tc.serviceTags)
 	}
 }
 
 func TestNonConfiguredService(t *testing.T) {
 	conf := &SDConfig{}
-	consulDiscovery, err := NewDiscovery(conf, nil, prometheus.NewRegistry())
-	if err != nil {
-		t.Errorf("Unexpected error when initializing discovery %v", err)
-	}
-	if !consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}) {
-		t.Errorf("Expected service %s to be watched", "nonConfiguredServiceName")
-	}
+
+	metrics := NewTestMetrics(t, conf, prometheus.NewRegistry())
+
+	consulDiscovery, err := NewDiscovery(conf, nil, metrics)
+	require.NoError(t, err, "when initializing discovery")
+	require.True(t, consulDiscovery.shouldWatch("nonConfiguredServiceName", []string{""}), "Expected service %s to be watched", "nonConfiguredServiceName")
 }
 
 const (
@@ -263,7 +271,10 @@ func newServer(t *testing.T) (*httptest.Server, *SDConfig) {
 
 func newDiscovery(t *testing.T, config *SDConfig) *Discovery {
 	logger := log.NewNopLogger()
-	d, err := NewDiscovery(config, logger, prometheus.NewRegistry())
+
+	metrics := NewTestMetrics(t, config, prometheus.NewRegistry())
+
+	d, err := NewDiscovery(config, logger, metrics)
 	require.NoError(t, err)
 	return d
 }
@@ -477,13 +488,10 @@ oauth2:
 			var config SDConfig
 			err := config.UnmarshalYAML(unmarshal([]byte(test.config)))
 			if err != nil {
-				require.Equalf(t, err.Error(), test.errMessage, "Expected error '%s', got '%v'", test.errMessage, err)
-				return
-			}
-			if test.errMessage != "" {
-				t.Errorf("Expected error %s, got none", test.errMessage)
+				require.EqualError(t, err, test.errMessage)
 				return
 			}
+			require.Empty(t, test.errMessage, "Expected error.")
 
 			require.Equal(t, test.expected, config)
 		})

discovery/consul/metrics.go

@@ -0,0 +1,73 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package consul
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*consulMetrics)(nil)
+
+type consulMetrics struct {
+	rpcFailuresCount prometheus.Counter
+	rpcDuration      *prometheus.SummaryVec
+
+	servicesRPCDuration prometheus.Observer
+	serviceRPCDuration  prometheus.Observer
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	m := &consulMetrics{
+		rpcFailuresCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Namespace: namespace,
+				Name:      "sd_consul_rpc_failures_total",
+				Help:      "The number of Consul RPC call failures.",
+			}),
+		rpcDuration: prometheus.NewSummaryVec(
+			prometheus.SummaryOpts{
+				Namespace:  namespace,
+				Name:       "sd_consul_rpc_duration_seconds",
+				Help:       "The duration of a Consul RPC call in seconds.",
+				Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
+			},
+			[]string{"endpoint", "call"},
+		),
+	}
+
+	m.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		m.rpcFailuresCount,
+		m.rpcDuration,
+	})
+
+	// Initialize metric vectors.
+	m.servicesRPCDuration = m.rpcDuration.WithLabelValues("catalog", "services")
+	m.serviceRPCDuration = m.rpcDuration.WithLabelValues("catalog", "service")
+
+	return m
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *consulMetrics) Register() error {
+	return m.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *consulMetrics) Unregister() {
+	m.metricRegisterer.UnregisterMetrics()
+}

discovery/digitalocean/digitalocean.go

@@ -63,6 +63,13 @@ func init() {
 	discovery.RegisterConfig(&SDConfig{})
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &digitaloceanMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // SDConfig is the configuration for DigitalOcean based service discovery.
 type SDConfig struct {
 	HTTPClientConfig config.HTTPClientConfig `yaml:",inline"`
@@ -76,7 +83,7 @@ func (*SDConfig) Name() string { return "digitalocean" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -104,7 +111,12 @@ type Discovery struct {
 }
 
 // NewDiscovery returns a new Discovery which periodically refreshes its targets.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*digitaloceanMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	d := &Discovery{
 		port: conf.Port,
 	}
@@ -127,11 +139,11 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "digitalocean",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "digitalocean",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 	return d, nil
@@ -165,7 +177,7 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 		}
 
 		labels := model.LabelSet{
-			doLabelID:          model.LabelValue(fmt.Sprintf("%d", droplet.ID)),
+			doLabelID:          model.LabelValue(strconv.Itoa(droplet.ID)),
 			doLabelName:        model.LabelValue(droplet.Name),
 			doLabelImage:       model.LabelValue(droplet.Image.Slug),
 			doLabelImageName:   model.LabelValue(droplet.Image.Name),

discovery/digitalocean/digitalocean_test.go

@@ -23,6 +23,8 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/common/model"
 	"github.com/stretchr/testify/require"
+
+	"github.com/prometheus/prometheus/discovery"
 )
 
 type DigitalOceanSDTestSuite struct {
@@ -47,7 +49,15 @@ func TestDigitalOceanSDRefresh(t *testing.T) {
 
 	cfg := DefaultSDConfig
 	cfg.HTTPClientConfig.BearerToken = tokenID
-	d, err := NewDiscovery(&cfg, log.NewNopLogger(), prometheus.NewRegistry())
+
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+	require.NoError(t, metrics.Register())
+	defer metrics.Unregister()
+	defer refreshMetrics.Unregister()
+
+	d, err := NewDiscovery(&cfg, log.NewNopLogger(), metrics)
 	require.NoError(t, err)
 	endpoint, err := url.Parse(sdmock.Mock.Endpoint())
 	require.NoError(t, err)

discovery/digitalocean/metrics.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package digitalocean
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*digitaloceanMetrics)(nil)
+
+type digitaloceanMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *digitaloceanMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *digitaloceanMetrics) Unregister() {}

discovery/discoverer_metrics_noop.go

@@ -0,0 +1,28 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package discovery
+
+// Create a dummy metrics struct, because this SD doesn't have any metrics.
+type NoopDiscovererMetrics struct{}
+
+var _ DiscovererMetrics = (*NoopDiscovererMetrics)(nil)
+
+// Register implements discovery.DiscovererMetrics.
+func (*NoopDiscovererMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (*NoopDiscovererMetrics) Unregister() {
+}

discovery/discovery.go

@@ -39,24 +39,47 @@ type Discoverer interface {
 	Run(ctx context.Context, up chan<- []*targetgroup.Group)
 }
 
+// Internal metrics of service discovery mechanisms.
+type DiscovererMetrics interface {
+	Register() error
+	Unregister()
+}
+
 // DiscovererOptions provides options for a Discoverer.
 type DiscovererOptions struct {
 	Logger log.Logger
 
-	// A registerer for the Discoverer's metrics.
-	// Some Discoverers may ignore this registerer and use the global one instead.
-	// For now this will work, because the Prometheus `main` function uses the global registry.
-	// However, in the future the Prometheus `main` function will be updated to not use the global registry.
-	// Hence, if a discoverer wants its metrics to be visible via the Prometheus executable's
-	// `/metrics` endpoint, it should use this explicit registerer.
-	// TODO(ptodev): Update this comment once the Prometheus `main` function does not use the global registry.
-	Registerer prometheus.Registerer
+	Metrics DiscovererMetrics
 
 	// Extra HTTP client options to expose to Discoverers. This field may be
 	// ignored; Discoverer implementations must opt-in to reading it.
 	HTTPClientOptions []config.HTTPClientOption
 }
 
+// Metrics used by the "refresh" package.
+// We define them here in the "discovery" package in order to avoid a cyclic dependency between
+// "discovery" and "refresh".
+type RefreshMetrics struct {
+	Failures prometheus.Counter
+	Duration prometheus.Observer
+}
+
+// Instantiate the metrics used by the "refresh" package.
+type RefreshMetricsInstantiator interface {
+	Instantiate(mech string) *RefreshMetrics
+}
+
+// An interface for registering, unregistering, and instantiating metrics for the "refresh" package.
+// Refresh metrics are registered and unregistered outside of the service discovery mechanism.
+// This is so that the same metrics can be reused across different service discovery mechanisms.
+// To manage refresh metrics inside the SD mechanism, we'd need to use const labels which are
+// specific to that SD. However, doing so would also expose too many unused metrics on
+// the Prometheus /metrics endpoint.
+type RefreshMetricsManager interface {
+	DiscovererMetrics
+	RefreshMetricsInstantiator
+}
+
 // A Config provides the configuration and constructor for a Discoverer.
 type Config interface {
 	// Name returns the name of the discovery mechanism.
@@ -65,6 +88,9 @@ type Config interface {
 	// NewDiscoverer returns a Discoverer for the Config
 	// with the given DiscovererOptions.
 	NewDiscoverer(DiscovererOptions) (Discoverer, error)
+
+	// NewDiscovererMetrics returns the metrics used by the service discovery.
+	NewDiscovererMetrics(prometheus.Registerer, RefreshMetricsInstantiator) DiscovererMetrics
 }
 
 // Configs is a slice of Config values that uses custom YAML marshaling and unmarshaling
@@ -119,6 +145,11 @@ func (c StaticConfig) NewDiscoverer(DiscovererOptions) (Discoverer, error) {
 	return staticDiscoverer(c), nil
 }
 
+// No metrics are needed for this service discovery mechanism.
+func (c StaticConfig) NewDiscovererMetrics(prometheus.Registerer, RefreshMetricsInstantiator) DiscovererMetrics {
+	return &NoopDiscovererMetrics{}
+}
+
 type staticDiscoverer []*targetgroup.Group
 
 func (c staticDiscoverer) Run(ctx context.Context, up chan<- []*targetgroup.Group) {

discovery/dns/dns.go

@@ -18,6 +18,7 @@ import (
 	"errors"
 	"fmt"
 	"net"
+	"strconv"
 	"strings"
 	"sync"
 	"time"
@@ -67,12 +68,17 @@ type SDConfig struct {
 	Port            int            `yaml:"port"` // Ignored for SRV records
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "dns" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(*c, opts.Logger, opts.Registerer)
+	return NewDiscovery(*c, opts.Logger, opts.Metrics)
 }
 
 // UnmarshalYAML implements the yaml.Unmarshaler interface.
@@ -102,18 +108,22 @@ func (c *SDConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 // the Discoverer interface.
 type Discovery struct {
 	*refresh.Discovery
-	names                    []string
-	port                     int
-	qtype                    uint16
-	logger                   log.Logger
-	dnsSDLookupsCount        prometheus.Counter
-	dnsSDLookupFailuresCount prometheus.Counter
+	names   []string
+	port    int
+	qtype   uint16
+	logger  log.Logger
+	metrics *dnsMetrics
 
 	lookupFn func(name string, qtype uint16, logger log.Logger) (*dns.Msg, error)
 }
 
 // NewDiscovery returns a new Discovery which periodically refreshes its targets.
-func NewDiscovery(conf SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*dnsMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -137,28 +147,16 @@ func NewDiscovery(conf SDConfig, logger log.Logger, reg prometheus.Registerer) (
 		port:     conf.Port,
 		logger:   logger,
 		lookupFn: lookupWithSearchPath,
-		dnsSDLookupsCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Namespace: namespace,
-				Name:      "sd_dns_lookups_total",
-				Help:      "The number of DNS-SD lookups.",
-			}),
-		dnsSDLookupFailuresCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Namespace: namespace,
-				Name:      "sd_dns_lookup_failures_total",
-				Help:      "The number of DNS-SD lookup failures.",
-			}),
+		metrics:  m,
 	}
 
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "dns",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: prometheus.NewRegistry(),
-			Metrics:  []prometheus.Collector{d.dnsSDLookupsCount, d.dnsSDLookupFailuresCount},
+			Logger:              logger,
+			Mech:                "dns",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 
@@ -195,15 +193,15 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 
 func (d *Discovery) refreshOne(ctx context.Context, name string, ch chan<- *targetgroup.Group) error {
 	response, err := d.lookupFn(name, d.qtype, d.logger)
-	d.dnsSDLookupsCount.Inc()
+	d.metrics.dnsSDLookupsCount.Inc()
 	if err != nil {
-		d.dnsSDLookupFailuresCount.Inc()
+		d.metrics.dnsSDLookupFailuresCount.Inc()
 		return err
 	}
 
 	tg := &targetgroup.Group{}
 	hostPort := func(a string, p int) model.LabelValue {
-		return model.LabelValue(net.JoinHostPort(a, fmt.Sprintf("%d", p)))
+		return model.LabelValue(net.JoinHostPort(a, strconv.Itoa(p)))
 	}
 
 	for _, record := range response.Answer {
@@ -212,7 +210,7 @@ func (d *Discovery) refreshOne(ctx context.Context, name string, ch chan<- *targ
 		switch addr := record.(type) {
 		case *dns.SRV:
 			dnsSrvRecordTarget = model.LabelValue(addr.Target)
-			dnsSrvRecordPort = model.LabelValue(fmt.Sprintf("%d", addr.Port))
+			dnsSrvRecordPort = model.LabelValue(strconv.Itoa(int(addr.Port)))
 
 			// Remove the final dot from rooted DNS names to make them look more usual.
 			addr.Target = strings.TrimRight(addr.Target, ".")

discovery/dns/dns_test.go

@@ -28,6 +28,7 @@ import (
 	"go.uber.org/goleak"
 	"gopkg.in/yaml.v2"
 
+	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
@@ -253,13 +254,21 @@ func TestDNS(t *testing.T) {
 		tc := tc
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
-			sd, err := NewDiscovery(tc.config, nil, prometheus.NewRegistry())
+
+			reg := prometheus.NewRegistry()
+			refreshMetrics := discovery.NewRefreshMetrics(reg)
+			metrics := tc.config.NewDiscovererMetrics(reg, refreshMetrics)
+			require.NoError(t, metrics.Register())
+
+			sd, err := NewDiscovery(tc.config, nil, metrics)
 			require.NoError(t, err)
 			sd.lookupFn = tc.lookup
 
 			tgs, err := sd.refresh(context.Background())
 			require.NoError(t, err)
 			require.Equal(t, tc.expected, tgs)
+
+			metrics.Unregister()
 		})
 	}
 }

discovery/dns/metrics.go

@@ -0,0 +1,66 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package dns
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*dnsMetrics)(nil)
+
+type dnsMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+
+	dnsSDLookupsCount        prometheus.Counter
+	dnsSDLookupFailuresCount prometheus.Counter
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	m := &dnsMetrics{
+		refreshMetrics: rmi,
+		dnsSDLookupsCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Namespace: namespace,
+				Name:      "sd_dns_lookups_total",
+				Help:      "The number of DNS-SD lookups.",
+			}),
+		dnsSDLookupFailuresCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Namespace: namespace,
+				Name:      "sd_dns_lookup_failures_total",
+				Help:      "The number of DNS-SD lookup failures.",
+			}),
+	}
+
+	m.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		m.dnsSDLookupsCount,
+		m.dnsSDLookupFailuresCount,
+	})
+
+	return m
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *dnsMetrics) Register() error {
+	return m.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *dnsMetrics) Unregister() {
+	m.metricRegisterer.UnregisterMetrics()
+}

discovery/eureka/client.go

@@ -81,7 +81,7 @@ const appListPath string = "/apps"
 func fetchApps(ctx context.Context, server string, client *http.Client) (*Applications, error) {
 	url := fmt.Sprintf("%s%s", server, appListPath)
 
-	request, err := http.NewRequest("GET", url, nil)
+	request, err := http.NewRequest(http.MethodGet, url, nil)
 	if err != nil {
 		return nil, err
 	}

discovery/eureka/eureka.go

@@ -16,6 +16,7 @@ package eureka
 import (
 	"context"
 	"errors"
+	"fmt"
 	"net"
 	"net/http"
 	"net/url"
@@ -76,12 +77,19 @@ type SDConfig struct {
 	RefreshInterval  model.Duration          `yaml:"refresh_interval,omitempty"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &eurekaMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "eureka" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -118,7 +126,12 @@ type Discovery struct {
 }
 
 // NewDiscovery creates a new Eureka discovery for the given role.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*eurekaMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	rt, err := config.NewRoundTripperFromConfig(conf.HTTPClientConfig, "eureka_sd")
 	if err != nil {
 		return nil, err
@@ -130,11 +143,11 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 	}
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "eureka",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "eureka",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 	return d, nil
@@ -215,7 +228,6 @@ func targetsForApp(app *Application) []model.LabelSet {
 		}
 
 		targets = append(targets, target)
-
 	}
 	return targets
 }

discovery/eureka/eureka_test.go

@@ -24,6 +24,7 @@ import (
 	"github.com/prometheus/common/model"
 	"github.com/stretchr/testify/require"
 
+	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
@@ -36,7 +37,17 @@ func testUpdateServices(respHandler http.HandlerFunc) ([]*targetgroup.Group, err
 		Server: ts.URL,
 	}
 
-	md, err := NewDiscovery(&conf, nil, prometheus.NewRegistry())
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	metrics := conf.NewDiscovererMetrics(reg, refreshMetrics)
+	err := metrics.Register()
+	if err != nil {
+		return nil, err
+	}
+	defer metrics.Unregister()
+	defer refreshMetrics.Unregister()
+
+	md, err := NewDiscovery(&conf, nil, metrics)
 	if err != nil {
 		return nil, err
 	}

discovery/eureka/metrics.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package eureka
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*eurekaMetrics)(nil)
+
+type eurekaMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *eurekaMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *eurekaMetrics) Unregister() {}

discovery/file/file.go

@@ -57,12 +57,17 @@ type SDConfig struct {
 	RefreshInterval model.Duration `yaml:"refresh_interval,omitempty"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "file" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -94,6 +99,9 @@ func (c *SDConfig) UnmarshalYAML(unmarshal func(interface{}) error) error {
 const fileSDFilepathLabel = model.MetaLabelPrefix + "filepath"
 
 // TimestampCollector is a Custom Collector for Timestamps of the files.
+// TODO(ptodev): Now that each file SD has its own TimestampCollector
+// inside discovery/file/metrics.go, we can refactor this collector
+// (or get rid of it) as each TimestampCollector instance will only use one discoverer.
 type TimestampCollector struct {
 	Description *prometheus.Desc
 	discoverers map[*Discovery]struct{}
@@ -169,16 +177,16 @@ type Discovery struct {
 	lastRefresh map[string]int
 	logger      log.Logger
 
-	fileSDReadErrorsCount  prometheus.Counter
-	fileSDScanDuration     prometheus.Summary
-	fileWatcherErrorsCount prometheus.Counter
-	fileSDTimeStamp        *TimestampCollector
-
-	metricRegisterer discovery.MetricRegisterer
+	metrics *fileMetrics
 }
 
 // NewDiscovery returns a new file discovery for the given paths.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	fm, ok := metrics.(*fileMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -188,33 +196,10 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 		interval:   time.Duration(conf.RefreshInterval),
 		timestamps: make(map[string]float64),
 		logger:     logger,
-		fileSDReadErrorsCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Name: "prometheus_sd_file_read_errors_total",
-				Help: "The number of File-SD read errors.",
-			}),
-		fileSDScanDuration: prometheus.NewSummary(
-			prometheus.SummaryOpts{
-				Name:       "prometheus_sd_file_scan_duration_seconds",
-				Help:       "The duration of the File-SD scan in seconds.",
-				Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
-			}),
-		fileWatcherErrorsCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Name: "prometheus_sd_file_watcher_errors_total",
-				Help: "The number of File-SD errors caused by filesystem watch failures.",
-			}),
-		fileSDTimeStamp: NewTimestampCollector(),
+		metrics:    fm,
 	}
 
-	disc.fileSDTimeStamp.addDiscoverer(disc)
-
-	disc.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
-		disc.fileSDReadErrorsCount,
-		disc.fileSDScanDuration,
-		disc.fileWatcherErrorsCount,
-		disc.fileSDTimeStamp,
-	})
+	fm.init(disc)
 
 	return disc, nil
 }
@@ -253,17 +238,10 @@ func (d *Discovery) watchFiles() {
 
 // Run implements the Discoverer interface.
 func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
-	err := d.metricRegisterer.RegisterMetrics()
-	if err != nil {
-		level.Error(d.logger).Log("msg", "Unable to register metrics", "err", err.Error())
-		return
-	}
-	defer d.metricRegisterer.UnregisterMetrics()
-
 	watcher, err := fsnotify.NewWatcher()
 	if err != nil {
 		level.Error(d.logger).Log("msg", "Error adding file watcher", "err", err)
-		d.fileWatcherErrorsCount.Inc()
+		d.metrics.fileWatcherErrorsCount.Inc()
 		return
 	}
 	d.watcher = watcher
@@ -327,7 +305,7 @@ func (d *Discovery) stop() {
 	done := make(chan struct{})
 	defer close(done)
 
-	d.fileSDTimeStamp.removeDiscoverer(d)
+	d.metrics.fileSDTimeStamp.removeDiscoverer(d)
 
 	// Closing the watcher will deadlock unless all events and errors are drained.
 	go func() {
@@ -353,13 +331,13 @@ func (d *Discovery) stop() {
 func (d *Discovery) refresh(ctx context.Context, ch chan<- []*targetgroup.Group) {
 	t0 := time.Now()
 	defer func() {
-		d.fileSDScanDuration.Observe(time.Since(t0).Seconds())
+		d.metrics.fileSDScanDuration.Observe(time.Since(t0).Seconds())
 	}()
 	ref := map[string]int{}
 	for _, p := range d.listFiles() {
 		tgroups, err := d.readFile(p)
 		if err != nil {
-			d.fileSDReadErrorsCount.Inc()
+			d.metrics.fileSDReadErrorsCount.Inc()
 
 			level.Error(d.logger).Log("msg", "Error reading file", "path", p, "err", err)
 			// Prevent deletion down below.

discovery/file/file_test.go

@@ -29,6 +29,7 @@ import (
 	"github.com/stretchr/testify/require"
 	"go.uber.org/goleak"
 
+	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
@@ -144,19 +145,28 @@ func (t *testRunner) run(files ...string) {
 	ctx, cancel := context.WithCancel(context.Background())
 	t.cancelSD = cancel
 	go func() {
+		conf := &SDConfig{
+			Files: files,
+			// Setting a high refresh interval to make sure that the tests only
+			// rely on file watches.
+			RefreshInterval: model.Duration(1 * time.Hour),
+		}
+
+		reg := prometheus.NewRegistry()
+		refreshMetrics := discovery.NewRefreshMetrics(reg)
+		metrics := conf.NewDiscovererMetrics(reg, refreshMetrics)
+		require.NoError(t, metrics.Register())
+
 		d, err := NewDiscovery(
-			&SDConfig{
-				Files: files,
-				// Setting a high refresh interval to make sure that the tests only
-				// rely on file watches.
-				RefreshInterval: model.Duration(1 * time.Hour),
-			},
+			conf,
 			nil,
-			prometheus.NewRegistry(),
+			metrics,
 		)
 		require.NoError(t, err)
 
 		d.Run(ctx, t.ch)
+
+		metrics.Unregister()
 	}()
 }
 
@@ -193,11 +203,11 @@ func (t *testRunner) targets() []*targetgroup.Group {
 func (t *testRunner) requireUpdate(ref time.Time, expected []*targetgroup.Group) {
 	t.Helper()
 
+	timeout := time.After(defaultWait)
 	for {
 		select {
-		case <-time.After(defaultWait):
+		case <-timeout:
 			t.Fatalf("Expected update but got none")
-			return
 		case <-time.After(defaultWait / 10):
 			if ref.Equal(t.lastReceive()) {
 				// No update received.
@@ -347,9 +357,7 @@ func TestInvalidFile(t *testing.T) {
 
 			// Verify that we've received nothing.
 			time.Sleep(defaultWait)
-			if runner.lastReceive().After(now) {
-				t.Fatalf("unexpected targets received: %v", runner.targets())
-			}
+			require.False(t, runner.lastReceive().After(now), "unexpected targets received: %v", runner.targets())
 		})
 	}
 }

discovery/file/metrics.go

@@ -0,0 +1,76 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package file
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*fileMetrics)(nil)
+
+type fileMetrics struct {
+	fileSDReadErrorsCount  prometheus.Counter
+	fileSDScanDuration     prometheus.Summary
+	fileWatcherErrorsCount prometheus.Counter
+	fileSDTimeStamp        *TimestampCollector
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	fm := &fileMetrics{
+		fileSDReadErrorsCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Name: "prometheus_sd_file_read_errors_total",
+				Help: "The number of File-SD read errors.",
+			}),
+		fileSDScanDuration: prometheus.NewSummary(
+			prometheus.SummaryOpts{
+				Name:       "prometheus_sd_file_scan_duration_seconds",
+				Help:       "The duration of the File-SD scan in seconds.",
+				Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
+			}),
+		fileWatcherErrorsCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Name: "prometheus_sd_file_watcher_errors_total",
+				Help: "The number of File-SD errors caused by filesystem watch failures.",
+			}),
+		fileSDTimeStamp: NewTimestampCollector(),
+	}
+
+	fm.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		fm.fileSDReadErrorsCount,
+		fm.fileSDScanDuration,
+		fm.fileWatcherErrorsCount,
+		fm.fileSDTimeStamp,
+	})
+
+	return fm
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (fm *fileMetrics) Register() error {
+	return fm.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (fm *fileMetrics) Unregister() {
+	fm.metricRegisterer.UnregisterMetrics()
+}
+
+func (fm *fileMetrics) init(disc *Discovery) {
+	fm.fileSDTimeStamp.addDiscoverer(disc)
+}

discovery/gce/gce.go

@@ -82,12 +82,19 @@ type SDConfig struct {
 	TagSeparator    string         `yaml:"tag_separator,omitempty"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &gceMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "gce" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(*c, opts.Logger, opts.Registerer)
+	return NewDiscovery(*c, opts.Logger, opts.Metrics)
 }
 
 // UnmarshalYAML implements the yaml.Unmarshaler interface.
@@ -122,7 +129,12 @@ type Discovery struct {
 }
 
 // NewDiscovery returns a new Discovery which periodically refreshes its targets.
-func NewDiscovery(conf SDConfig, logger log.Logger, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*gceMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	d := &Discovery{
 		project:      conf.Project,
 		zone:         conf.Zone,
@@ -143,11 +155,11 @@ func NewDiscovery(conf SDConfig, logger log.Logger, reg prometheus.Registerer) (
 
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "gce",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "gce",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 	return d, nil

discovery/gce/metrics.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package gce
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*gceMetrics)(nil)
+
+type gceMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *gceMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *gceMetrics) Unregister() {}

discovery/hetzner/hcloud.go

@@ -15,7 +15,6 @@ package hetzner
 
 import (
 	"context"
-	"fmt"
 	"net"
 	"net/http"
 	"strconv"
@@ -92,7 +91,7 @@ func (d *hcloudDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, er
 	for i, server := range servers {
 		labels := model.LabelSet{
 			hetznerLabelRole:              model.LabelValue(HetznerRoleHcloud),
-			hetznerLabelServerID:          model.LabelValue(fmt.Sprintf("%d", server.ID)),
+			hetznerLabelServerID:          model.LabelValue(strconv.FormatInt(server.ID, 10)),
 			hetznerLabelServerName:        model.LabelValue(server.Name),
 			hetznerLabelDatacenter:        model.LabelValue(server.Datacenter.Name),
 			hetznerLabelPublicIPv4:        model.LabelValue(server.PublicNet.IPv4.IP.String()),
@@ -102,10 +101,10 @@ func (d *hcloudDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, er
 			hetznerLabelHcloudDatacenterLocation:            model.LabelValue(server.Datacenter.Location.Name),
 			hetznerLabelHcloudDatacenterLocationNetworkZone: model.LabelValue(server.Datacenter.Location.NetworkZone),
 			hetznerLabelHcloudType:                          model.LabelValue(server.ServerType.Name),
-			hetznerLabelHcloudCPUCores:                      model.LabelValue(fmt.Sprintf("%d", server.ServerType.Cores)),
+			hetznerLabelHcloudCPUCores:                      model.LabelValue(strconv.Itoa(server.ServerType.Cores)),
 			hetznerLabelHcloudCPUType:                       model.LabelValue(server.ServerType.CPUType),
-			hetznerLabelHcloudMemoryGB:                      model.LabelValue(fmt.Sprintf("%d", int(server.ServerType.Memory))),
-			hetznerLabelHcloudDiskGB:                        model.LabelValue(fmt.Sprintf("%d", server.ServerType.Disk)),
+			hetznerLabelHcloudMemoryGB:                      model.LabelValue(strconv.Itoa(int(server.ServerType.Memory))),
+			hetznerLabelHcloudDiskGB:                        model.LabelValue(strconv.Itoa(server.ServerType.Disk)),
 
 			model.AddressLabel: model.LabelValue(net.JoinHostPort(server.PublicNet.IPv4.IP.String(), strconv.FormatUint(uint64(d.port), 10))),
 		}

discovery/hetzner/hetzner.go

@@ -63,12 +63,19 @@ type SDConfig struct {
 	robotEndpoint   string         // For tests only.
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &hetznerMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "hetzner" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.Metrics)
 }
 
 type refresher interface {
@@ -128,7 +135,12 @@ type Discovery struct {
 }
 
 // NewDiscovery returns a new Discovery which periodically refreshes its targets.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*refresh.Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*refresh.Discovery, error) {
+	m, ok := metrics.(*hetznerMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	r, err := newRefresher(conf, logger)
 	if err != nil {
 		return nil, err
@@ -136,11 +148,11 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 
 	return refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "hetzner",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: r.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "hetzner",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            r.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	), nil
 }

discovery/hetzner/metrics.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package hetzner
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*hetznerMetrics)(nil)
+
+type hetznerMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *hetznerMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *hetznerMetrics) Unregister() {}

discovery/hetzner/robot.go

@@ -70,7 +70,7 @@ func newRobotDiscovery(conf *SDConfig, _ log.Logger) (*robotDiscovery, error) {
 }
 
 func (d *robotDiscovery) refresh(context.Context) ([]*targetgroup.Group, error) {
-	req, err := http.NewRequest("GET", d.endpoint+"/server", nil)
+	req, err := http.NewRequest(http.MethodGet, d.endpoint+"/server", nil)
 	if err != nil {
 		return nil, err
 	}
@@ -112,7 +112,7 @@ func (d *robotDiscovery) refresh(context.Context) ([]*targetgroup.Group, error)
 			hetznerLabelPublicIPv4:     model.LabelValue(server.Server.ServerIP),
 			hetznerLabelServerStatus:   model.LabelValue(server.Server.Status),
 			hetznerLabelRobotProduct:   model.LabelValue(server.Server.Product),
-			hetznerLabelRobotCancelled: model.LabelValue(fmt.Sprintf("%t", server.Server.Canceled)),
+			hetznerLabelRobotCancelled: model.LabelValue(strconv.FormatBool(server.Server.Canceled)),
 
 			model.AddressLabel: model.LabelValue(net.JoinHostPort(server.Server.ServerIP, strconv.FormatUint(uint64(d.port), 10))),
 		}
@@ -122,7 +122,6 @@ func (d *robotDiscovery) refresh(context.Context) ([]*targetgroup.Group, error)
 				labels[hetznerLabelPublicIPv6Network] = model.LabelValue(fmt.Sprintf("%s/%s", subnet.IP, subnet.Mask))
 				break
 			}
-
 		}
 		targets[i] = labels
 	}

discovery/http/http.go

@@ -58,12 +58,17 @@ type SDConfig struct {
 	URL              string                  `yaml:"url"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "http" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(c, opts.Logger, opts.HTTPClientOptions, opts.Registerer)
+	return NewDiscovery(c, opts.Logger, opts.HTTPClientOptions, opts.Metrics)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -105,11 +110,16 @@ type Discovery struct {
 	client          *http.Client
 	refreshInterval time.Duration
 	tgLastLength    int
-	failuresCount   prometheus.Counter
+	metrics         *httpMetrics
 }
 
 // NewDiscovery returns a new HTTP discovery for the given config.
-func NewDiscovery(conf *SDConfig, logger log.Logger, clientOpts []config.HTTPClientOption, reg prometheus.Registerer) (*Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, clientOpts []config.HTTPClientOption, metrics discovery.DiscovererMetrics) (*Discovery, error) {
+	m, ok := metrics.(*httpMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -124,28 +134,23 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, clientOpts []config.HTTPCli
 		url:             conf.URL,
 		client:          client,
 		refreshInterval: time.Duration(conf.RefreshInterval), // Stored to be sent as headers.
-		failuresCount: prometheus.NewCounter(
-			prometheus.CounterOpts{
-				Name: "prometheus_sd_http_failures_total",
-				Help: "Number of HTTP service discovery refresh failures.",
-			}),
+		metrics:         m,
 	}
 
 	d.Discovery = refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "http",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.Refresh,
-			Registry: reg,
-			Metrics:  []prometheus.Collector{d.failuresCount},
+			Logger:              logger,
+			Mech:                "http",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.Refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	)
 	return d, nil
 }
 
 func (d *Discovery) Refresh(ctx context.Context) ([]*targetgroup.Group, error) {
-	req, err := http.NewRequest("GET", d.url, nil)
+	req, err := http.NewRequest(http.MethodGet, d.url, nil)
 	if err != nil {
 		return nil, err
 	}
@@ -155,7 +160,7 @@ func (d *Discovery) Refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 
 	resp, err := d.client.Do(req.WithContext(ctx))
 	if err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, err
 	}
 	defer func() {
@@ -164,31 +169,31 @@ func (d *Discovery) Refresh(ctx context.Context) ([]*targetgroup.Group, error) {
 	}()
 
 	if resp.StatusCode != http.StatusOK {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, fmt.Errorf("server returned HTTP status %s", resp.Status)
 	}
 
 	if !matchContentType.MatchString(strings.TrimSpace(resp.Header.Get("Content-Type"))) {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, fmt.Errorf("unsupported content type %q", resp.Header.Get("Content-Type"))
 	}
 
 	b, err := io.ReadAll(resp.Body)
 	if err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, err
 	}
 
 	var targetGroups []*targetgroup.Group
 
 	if err := json.Unmarshal(b, &targetGroups); err != nil {
-		d.failuresCount.Inc()
+		d.metrics.failuresCount.Inc()
 		return nil, err
 	}
 
 	for i, tg := range targetGroups {
 		if tg == nil {
-			d.failuresCount.Inc()
+			d.metrics.failuresCount.Inc()
 			err = errors.New("nil target group item found")
 			return nil, err
 		}

discovery/http/http_test.go

@@ -28,6 +28,7 @@ import (
 	"github.com/prometheus/common/model"
 	"github.com/stretchr/testify/require"
 
+	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
@@ -41,7 +42,14 @@ func TestHTTPValidRefresh(t *testing.T) {
 		RefreshInterval:  model.Duration(30 * time.Second),
 	}
 
-	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, prometheus.NewRegistry())
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	defer refreshMetrics.Unregister()
+	metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+	require.NoError(t, metrics.Register())
+	defer metrics.Unregister()
+
+	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, metrics)
 	require.NoError(t, err)
 
 	ctx := context.Background()
@@ -63,7 +71,7 @@ func TestHTTPValidRefresh(t *testing.T) {
 		},
 	}
 	require.Equal(t, expectedTargets, tgs)
-	require.Equal(t, 0.0, getFailureCount(d.failuresCount))
+	require.Equal(t, 0.0, getFailureCount(d.metrics.failuresCount))
 }
 
 func TestHTTPInvalidCode(t *testing.T) {
@@ -79,13 +87,20 @@ func TestHTTPInvalidCode(t *testing.T) {
 		RefreshInterval:  model.Duration(30 * time.Second),
 	}
 
-	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, prometheus.NewRegistry())
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	defer refreshMetrics.Unregister()
+	metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+	require.NoError(t, metrics.Register())
+	defer metrics.Unregister()
+
+	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, metrics)
 	require.NoError(t, err)
 
 	ctx := context.Background()
 	_, err = d.Refresh(ctx)
 	require.EqualError(t, err, "server returned HTTP status 400 Bad Request")
-	require.Equal(t, 1.0, getFailureCount(d.failuresCount))
+	require.Equal(t, 1.0, getFailureCount(d.metrics.failuresCount))
 }
 
 func TestHTTPInvalidFormat(t *testing.T) {
@@ -101,13 +116,20 @@ func TestHTTPInvalidFormat(t *testing.T) {
 		RefreshInterval:  model.Duration(30 * time.Second),
 	}
 
-	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, prometheus.NewRegistry())
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	defer refreshMetrics.Unregister()
+	metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+	require.NoError(t, metrics.Register())
+	defer metrics.Unregister()
+
+	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, metrics)
 	require.NoError(t, err)
 
 	ctx := context.Background()
 	_, err = d.Refresh(ctx)
 	require.EqualError(t, err, `unsupported content type "text/plain; charset=utf-8"`)
-	require.Equal(t, 1.0, getFailureCount(d.failuresCount))
+	require.Equal(t, 1.0, getFailureCount(d.metrics.failuresCount))
 }
 
 func getFailureCount(failuresCount prometheus.Counter) float64 {
@@ -412,7 +434,15 @@ func TestSourceDisappeared(t *testing.T) {
 		URL:              ts.URL,
 		RefreshInterval:  model.Duration(1 * time.Second),
 	}
-	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, prometheus.NewRegistry())
+
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	defer refreshMetrics.Unregister()
+	metrics := cfg.NewDiscovererMetrics(reg, refreshMetrics)
+	require.NoError(t, metrics.Register())
+	defer metrics.Unregister()
+
+	d, err := NewDiscovery(&cfg, log.NewNopLogger(), nil, metrics)
 	require.NoError(t, err)
 	for _, test := range cases {
 		ctx := context.Background()

discovery/http/metrics.go

@@ -0,0 +1,57 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package http
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*httpMetrics)(nil)
+
+type httpMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+
+	failuresCount prometheus.Counter
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	m := &httpMetrics{
+		refreshMetrics: rmi,
+		failuresCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Name: "prometheus_sd_http_failures_total",
+				Help: "Number of HTTP service discovery refresh failures.",
+			}),
+	}
+
+	m.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		m.failuresCount,
+	})
+
+	return m
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *httpMetrics) Register() error {
+	return m.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *httpMetrics) Unregister() {
+	m.metricRegisterer.UnregisterMetrics()
+}

discovery/ionos/ionos.go

@@ -15,16 +15,16 @@ package ionos
 
 import (
 	"errors"
+	"fmt"
 	"time"
 
 	"github.com/go-kit/log"
+	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/common/config"
 	"github.com/prometheus/common/model"
 
 	"github.com/prometheus/prometheus/discovery"
 	"github.com/prometheus/prometheus/discovery/refresh"
-
-	"github.com/prometheus/client_golang/prometheus"
 )
 
 const (
@@ -43,7 +43,12 @@ func init() {
 type Discovery struct{}
 
 // NewDiscovery returns a new refresh.Discovery for IONOS Cloud.
-func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer) (*refresh.Discovery, error) {
+func NewDiscovery(conf *SDConfig, logger log.Logger, metrics discovery.DiscovererMetrics) (*refresh.Discovery, error) {
+	m, ok := metrics.(*ionosMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if conf.ionosEndpoint == "" {
 		conf.ionosEndpoint = "https://api.ionos.com"
 	}
@@ -55,11 +60,11 @@ func NewDiscovery(conf *SDConfig, logger log.Logger, reg prometheus.Registerer)
 
 	return refresh.NewDiscovery(
 		refresh.Options{
-			Logger:   logger,
-			Mech:     "ionos",
-			Interval: time.Duration(conf.RefreshInterval),
-			RefreshF: d.refresh,
-			Registry: reg,
+			Logger:              logger,
+			Mech:                "ionos",
+			Interval:            time.Duration(conf.RefreshInterval),
+			RefreshF:            d.refresh,
+			MetricsInstantiator: m.refreshMetrics,
 		},
 	), nil
 }
@@ -84,6 +89,13 @@ type SDConfig struct {
 	ionosEndpoint string // For tests only.
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return &ionosMetrics{
+		refreshMetrics: rmi,
+	}
+}
+
 // Name returns the name of the IONOS Cloud service discovery.
 func (c SDConfig) Name() string {
 	return "ionos"
@@ -91,7 +103,7 @@ func (c SDConfig) Name() string {
 
 // NewDiscoverer returns a new discovery.Discoverer for IONOS Cloud.
 func (c SDConfig) NewDiscoverer(options discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return NewDiscovery(&c, options.Logger, options.Registerer)
+	return NewDiscovery(&c, options.Logger, options.Metrics)
 }
 
 // UnmarshalYAML implements the yaml.Unmarshaler interface.

discovery/ionos/metrics.go

@@ -0,0 +1,32 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package ionos
+
+import (
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*ionosMetrics)(nil)
+
+type ionosMetrics struct {
+	refreshMetrics discovery.RefreshMetricsInstantiator
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *ionosMetrics) Register() error {
+	return nil
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *ionosMetrics) Unregister() {}

discovery/kubernetes/endpoints.go

@@ -62,6 +62,8 @@ func NewEndpoints(l log.Logger, eps cache.SharedIndexInformer, svc, pod, node ca
 	svcUpdateCount := eventCount.WithLabelValues(RoleService.String(), MetricLabelRoleUpdate)
 	svcDeleteCount := eventCount.WithLabelValues(RoleService.String(), MetricLabelRoleDelete)
 
+	podUpdateCount := eventCount.WithLabelValues(RolePod.String(), MetricLabelRoleUpdate)
+
 	e := &Endpoints{
 		logger:           l,
 		endpointsInf:     eps,
@@ -131,6 +133,29 @@ func NewEndpoints(l log.Logger, eps cache.SharedIndexInformer, svc, pod, node ca
 	if err != nil {
 		level.Error(l).Log("msg", "Error adding services event handler.", "err", err)
 	}
+	_, err = e.podInf.AddEventHandler(cache.ResourceEventHandlerFuncs{
+		UpdateFunc: func(old, cur interface{}) {
+			podUpdateCount.Inc()
+			oldPod, ok := old.(*apiv1.Pod)
+			if !ok {
+				return
+			}
+
+			curPod, ok := cur.(*apiv1.Pod)
+			if !ok {
+				return
+			}
+
+			// the Pod's phase may change without triggering an update on the Endpoints/Service.
+			// https://github.com/prometheus/prometheus/issues/11305.
+			if curPod.Status.Phase != oldPod.Status.Phase {
+				e.enqueuePod(namespacedName(curPod.Namespace, curPod.Name))
+			}
+		},
+	})
+	if err != nil {
+		level.Error(l).Log("msg", "Error adding pods event handler.", "err", err)
+	}
 	if e.withNodeMetadata {
 		_, err = e.nodeInf.AddEventHandler(cache.ResourceEventHandlerFuncs{
 			AddFunc: func(o interface{}) {
@@ -166,6 +191,18 @@ func (e *Endpoints) enqueueNode(nodeName string) {
 	}
 }
 
+func (e *Endpoints) enqueuePod(podNamespacedName string) {
+	endpoints, err := e.endpointsInf.GetIndexer().ByIndex(podIndex, podNamespacedName)
+	if err != nil {
+		level.Error(e.logger).Log("msg", "Error getting endpoints for pod", "pod", podNamespacedName, "err", err)
+		return
+	}
+
+	for _, endpoint := range endpoints {
+		e.enqueue(endpoint)
+	}
+}
+
 func (e *Endpoints) enqueue(obj interface{}) {
 	key, err := cache.DeletionHandlingMetaNamespaceKeyFunc(obj)
 	if err != nil {
@@ -312,7 +349,7 @@ func (e *Endpoints) buildEndpoints(eps *apiv1.Endpoints) *targetgroup.Group {
 			tg.Targets = append(tg.Targets, target)
 			return
 		}
-		s := pod.Namespace + "/" + pod.Name
+		s := namespacedName(pod.Namespace, pod.Name)
 
 		sp, ok := seenPods[s]
 		if !ok {
@@ -369,6 +406,11 @@ func (e *Endpoints) buildEndpoints(eps *apiv1.Endpoints) *targetgroup.Group {
 	// For all seen pods, check all container ports. If they were not covered
 	// by one of the service endpoints, generate targets for them.
 	for _, pe := range seenPods {
+		// PodIP can be empty when a pod is starting or has been evicted.
+		if len(pe.pod.Status.PodIP) == 0 {
+			continue
+		}
+
 		for _, c := range pe.pod.Spec.Containers {
 			for _, cport := range c.Ports {
 				hasSeenPort := func() bool {
@@ -383,21 +425,18 @@ func (e *Endpoints) buildEndpoints(eps *apiv1.Endpoints) *targetgroup.Group {
 					continue
 				}
 
-				// PodIP can be empty when a pod is starting or has been evicted.
-				if len(pe.pod.Status.PodIP) != 0 {
-					a := net.JoinHostPort(pe.pod.Status.PodIP, strconv.FormatUint(uint64(cport.ContainerPort), 10))
-					ports := strconv.FormatUint(uint64(cport.ContainerPort), 10)
+				a := net.JoinHostPort(pe.pod.Status.PodIP, strconv.FormatUint(uint64(cport.ContainerPort), 10))
+				ports := strconv.FormatUint(uint64(cport.ContainerPort), 10)
 
-					target := model.LabelSet{
-						model.AddressLabel:            lv(a),
-						podContainerNameLabel:         lv(c.Name),
-						podContainerImageLabel:        lv(c.Image),
-						podContainerPortNameLabel:     lv(cport.Name),
-						podContainerPortNumberLabel:   lv(ports),
-						podContainerPortProtocolLabel: lv(string(cport.Protocol)),
-					}
-					tg.Targets = append(tg.Targets, target.Merge(podLabels(pe.pod)))
+				target := model.LabelSet{
+					model.AddressLabel:            lv(a),
+					podContainerNameLabel:         lv(c.Name),
+					podContainerImageLabel:        lv(c.Image),
+					podContainerPortNameLabel:     lv(cport.Name),
+					podContainerPortNumberLabel:   lv(ports),
+					podContainerPortProtocolLabel: lv(string(cport.Protocol)),
 				}
+				tg.Targets = append(tg.Targets, target.Merge(podLabels(pe.pod)))
 			}
 		}
 	}

discovery/kubernetes/endpoints_test.go

@@ -969,3 +969,123 @@ func TestEndpointsDiscoveryEmptyPodStatus(t *testing.T) {
 		expectedRes:      map[string]*targetgroup.Group{},
 	}.Run(t)
 }
+
+// TestEndpointsUpdatePod makes sure that Endpoints discovery detects underlying Pods changes.
+// See https://github.com/prometheus/prometheus/issues/11305 for more details.
+func TestEndpointsDiscoveryUpdatePod(t *testing.T) {
+	pod := &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "testpod",
+			Namespace: "default",
+			UID:       types.UID("deadbeef"),
+		},
+		Spec: v1.PodSpec{
+			NodeName: "testnode",
+			Containers: []v1.Container{
+				{
+					Name:  "c1",
+					Image: "c1:latest",
+					Ports: []v1.ContainerPort{
+						{
+							Name:          "mainport",
+							ContainerPort: 9000,
+							Protocol:      v1.ProtocolTCP,
+						},
+					},
+				},
+			},
+		},
+		Status: v1.PodStatus{
+			// Pod is in Pending phase when discovered for first time.
+			Phase: "Pending",
+			Conditions: []v1.PodCondition{
+				{
+					Type:   v1.PodReady,
+					Status: v1.ConditionFalse,
+				},
+			},
+			HostIP: "2.3.4.5",
+			PodIP:  "4.3.2.1",
+		},
+	}
+	objs := []runtime.Object{
+		&v1.Endpoints{
+			ObjectMeta: metav1.ObjectMeta{
+				Name:      "testendpoints",
+				Namespace: "default",
+			},
+			Subsets: []v1.EndpointSubset{
+				{
+					Addresses: []v1.EndpointAddress{
+						{
+							IP: "4.3.2.1",
+							// The Pending Pod may be included because the Endpoints was created manually.
+							// Or because the corresponding service has ".spec.publishNotReadyAddresses: true".
+							TargetRef: &v1.ObjectReference{
+								Kind:      "Pod",
+								Name:      "testpod",
+								Namespace: "default",
+							},
+						},
+					},
+					Ports: []v1.EndpointPort{
+						{
+							Name:     "mainport",
+							Port:     9000,
+							Protocol: v1.ProtocolTCP,
+						},
+					},
+				},
+			},
+		},
+		pod,
+	}
+	n, c := makeDiscovery(RoleEndpoint, NamespaceDiscovery{}, objs...)
+
+	k8sDiscoveryTest{
+		discovery: n,
+		afterStart: func() {
+			// the Pod becomes Ready.
+			pod.Status.Phase = "Running"
+			pod.Status.Conditions = []v1.PodCondition{
+				{
+					Type:   v1.PodReady,
+					Status: v1.ConditionTrue,
+				},
+			}
+			c.CoreV1().Pods(pod.Namespace).Update(context.Background(), pod, metav1.UpdateOptions{})
+		},
+		expectedMaxItems: 2,
+		expectedRes: map[string]*targetgroup.Group{
+			"endpoints/default/testendpoints": {
+				Targets: []model.LabelSet{
+					{
+						"__address__":                                    "4.3.2.1:9000",
+						"__meta_kubernetes_endpoint_port_name":           "mainport",
+						"__meta_kubernetes_endpoint_port_protocol":       "TCP",
+						"__meta_kubernetes_endpoint_ready":               "true",
+						"__meta_kubernetes_endpoint_address_target_kind": "Pod",
+						"__meta_kubernetes_endpoint_address_target_name": "testpod",
+						"__meta_kubernetes_pod_name":                     "testpod",
+						"__meta_kubernetes_pod_ip":                       "4.3.2.1",
+						"__meta_kubernetes_pod_ready":                    "true",
+						"__meta_kubernetes_pod_phase":                    "Running",
+						"__meta_kubernetes_pod_node_name":                "testnode",
+						"__meta_kubernetes_pod_host_ip":                  "2.3.4.5",
+						"__meta_kubernetes_pod_container_name":           "c1",
+						"__meta_kubernetes_pod_container_image":          "c1:latest",
+						"__meta_kubernetes_pod_container_port_name":      "mainport",
+						"__meta_kubernetes_pod_container_port_number":    "9000",
+						"__meta_kubernetes_pod_container_port_protocol":  "TCP",
+						"__meta_kubernetes_pod_uid":                      "deadbeef",
+					},
+				},
+				Labels: model.LabelSet{
+					"__meta_kubernetes_namespace":      "default",
+					"__meta_kubernetes_endpoints_name": "testendpoints",
+				},
+				Source: "endpoints/default/testendpoints",
+			},
+		},
+	}.Run(t)
+}

discovery/kubernetes/endpointslice.go

@@ -265,7 +265,9 @@ const (
 	endpointSliceEndpointConditionsReadyLabel       = metaLabelPrefix + "endpointslice_endpoint_conditions_ready"
 	endpointSliceEndpointConditionsServingLabel     = metaLabelPrefix + "endpointslice_endpoint_conditions_serving"
 	endpointSliceEndpointConditionsTerminatingLabel = metaLabelPrefix + "endpointslice_endpoint_conditions_terminating"
+	endpointSliceEndpointZoneLabel                  = metaLabelPrefix + "endpointslice_endpoint_zone"
 	endpointSliceEndpointHostnameLabel              = metaLabelPrefix + "endpointslice_endpoint_hostname"
+	endpointSliceEndpointNodenameLabel              = metaLabelPrefix + "endpointslice_endpoint_node_name"
 	endpointSliceAddressTargetKindLabel             = metaLabelPrefix + "endpointslice_address_target_kind"
 	endpointSliceAddressTargetNameLabel             = metaLabelPrefix + "endpointslice_address_target_name"
 	endpointSliceEndpointTopologyLabelPrefix        = metaLabelPrefix + "endpointslice_endpoint_topology_"
@@ -338,6 +340,14 @@ func (e *EndpointSlice) buildEndpointSlice(eps endpointSliceAdaptor) *targetgrou
 			target[model.LabelName(endpointSliceAddressTargetNameLabel)] = lv(ep.targetRef().Name)
 		}
 
+		if ep.nodename() != nil {
+			target[endpointSliceEndpointNodenameLabel] = lv(*ep.nodename())
+		}
+
+		if ep.zone() != nil {
+			target[model.LabelName(endpointSliceEndpointZoneLabel)] = lv(*ep.zone())
+		}
+
 		for k, v := range ep.topology() {
 			ln := strutil.SanitizeLabelName(k)
 			target[model.LabelName(endpointSliceEndpointTopologyLabelPrefix+ln)] = lv(v)
@@ -358,7 +368,7 @@ func (e *EndpointSlice) buildEndpointSlice(eps endpointSliceAdaptor) *targetgrou
 			tg.Targets = append(tg.Targets, target)
 			return
 		}
-		s := pod.Namespace + "/" + pod.Name
+		s := namespacedName(pod.Namespace, pod.Name)
 
 		sp, ok := seenPods[s]
 		if !ok {
@@ -405,6 +415,11 @@ func (e *EndpointSlice) buildEndpointSlice(eps endpointSliceAdaptor) *targetgrou
 	// For all seen pods, check all container ports. If they were not covered
 	// by one of the service endpoints, generate targets for them.
 	for _, pe := range seenPods {
+		// PodIP can be empty when a pod is starting or has been evicted.
+		if len(pe.pod.Status.PodIP) == 0 {
+			continue
+		}
+
 		for _, c := range pe.pod.Spec.Containers {
 			for _, cport := range c.Ports {
 				hasSeenPort := func() bool {
@@ -422,21 +437,18 @@ func (e *EndpointSlice) buildEndpointSlice(eps endpointSliceAdaptor) *targetgrou
 					continue
 				}
 
-				// PodIP can be empty when a pod is starting or has been evicted.
-				if len(pe.pod.Status.PodIP) != 0 {
-					a := net.JoinHostPort(pe.pod.Status.PodIP, strconv.FormatUint(uint64(cport.ContainerPort), 10))
-					ports := strconv.FormatUint(uint64(cport.ContainerPort), 10)
+				a := net.JoinHostPort(pe.pod.Status.PodIP, strconv.FormatUint(uint64(cport.ContainerPort), 10))
+				ports := strconv.FormatUint(uint64(cport.ContainerPort), 10)
 
-					target := model.LabelSet{
-						model.AddressLabel:            lv(a),
-						podContainerNameLabel:         lv(c.Name),
-						podContainerImageLabel:        lv(c.Image),
-						podContainerPortNameLabel:     lv(cport.Name),
-						podContainerPortNumberLabel:   lv(ports),
-						podContainerPortProtocolLabel: lv(string(cport.Protocol)),
-					}
-					tg.Targets = append(tg.Targets, target.Merge(podLabels(pe.pod)))
+				target := model.LabelSet{
+					model.AddressLabel:            lv(a),
+					podContainerNameLabel:         lv(c.Name),
+					podContainerImageLabel:        lv(c.Image),
+					podContainerPortNameLabel:     lv(cport.Name),
+					podContainerPortNumberLabel:   lv(ports),
+					podContainerPortProtocolLabel: lv(string(cport.Protocol)),
 				}
+				tg.Targets = append(tg.Targets, target.Merge(podLabels(pe.pod)))
 			}
 		}
 	}

discovery/kubernetes/endpointslice_adaptor.go

@@ -44,6 +44,7 @@ type endpointSliceEndpointAdaptor interface {
 	addresses() []string
 	hostname() *string
 	nodename() *string
+	zone() *string
 	conditions() endpointSliceEndpointConditionsAdaptor
 	targetRef() *corev1.ObjectReference
 	topology() map[string]string
@@ -181,6 +182,10 @@ func (e *endpointSliceEndpointAdaptorV1) nodename() *string {
 	return e.endpoint.NodeName
 }
 
+func (e *endpointSliceEndpointAdaptorV1) zone() *string {
+	return e.endpoint.Zone
+}
+
 func (e *endpointSliceEndpointAdaptorV1) conditions() endpointSliceEndpointConditionsAdaptor {
 	return newEndpointSliceEndpointConditionsAdaptorFromV1(e.endpoint.Conditions)
 }
@@ -233,6 +238,10 @@ func (e *endpointSliceEndpointAdaptorV1beta1) nodename() *string {
 	return e.endpoint.NodeName
 }
 
+func (e *endpointSliceEndpointAdaptorV1beta1) zone() *string {
+	return nil
+}
+
 func (e *endpointSliceEndpointAdaptorV1beta1) conditions() endpointSliceEndpointConditionsAdaptor {
 	return newEndpointSliceEndpointConditionsAdaptorFromV1beta1(e.endpoint.Conditions)
 }

discovery/kubernetes/endpointslice_test.go

@@ -18,6 +18,7 @@ import (
 	"testing"
 
 	"github.com/prometheus/common/model"
+	"github.com/stretchr/testify/require"
 	corev1 "k8s.io/api/core/v1"
 	v1 "k8s.io/api/discovery/v1"
 	"k8s.io/api/discovery/v1beta1"
@@ -79,6 +80,7 @@ func makeEndpointSliceV1() *v1.EndpointSlice {
 				DeprecatedTopology: map[string]string{
 					"topology": "value",
 				},
+				Zone: strptr("us-east-1a"),
 			}, {
 				Addresses: []string{"2.3.4.5"},
 				Conditions: v1.EndpointConditions{
@@ -86,6 +88,7 @@ func makeEndpointSliceV1() *v1.EndpointSlice {
 					Serving:     boolptr(true),
 					Terminating: boolptr(false),
 				},
+				Zone: strptr("us-east-1b"),
 			}, {
 				Addresses: []string{"3.4.5.6"},
 				Conditions: v1.EndpointConditions{
@@ -93,6 +96,7 @@ func makeEndpointSliceV1() *v1.EndpointSlice {
 					Serving:     boolptr(true),
 					Terminating: boolptr(true),
 				},
+				Zone: strptr("us-east-1c"),
 			}, {
 				Addresses: []string{"4.5.6.7"},
 				Conditions: v1.EndpointConditions{
@@ -104,6 +108,7 @@ func makeEndpointSliceV1() *v1.EndpointSlice {
 					Kind: "Node",
 					Name: "barbaz",
 				},
+				Zone: strptr("us-east-1a"),
 			},
 		},
 	}
@@ -184,8 +189,10 @@ func TestEndpointSliceDiscoveryBeforeRun(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -196,6 +203,7 @@ func TestEndpointSliceDiscoveryBeforeRun(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -206,6 +214,7 @@ func TestEndpointSliceDiscoveryBeforeRun(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -218,6 +227,7 @@ func TestEndpointSliceDiscoveryBeforeRun(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -451,8 +461,10 @@ func TestEndpointSliceDiscoveryDelete(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -463,6 +475,7 @@ func TestEndpointSliceDiscoveryDelete(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -473,6 +486,7 @@ func TestEndpointSliceDiscoveryDelete(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -485,6 +499,7 @@ func TestEndpointSliceDiscoveryDelete(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -551,8 +566,10 @@ func TestEndpointSliceDiscoveryUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -563,6 +580,7 @@ func TestEndpointSliceDiscoveryUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -573,6 +591,7 @@ func TestEndpointSliceDiscoveryUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -585,6 +604,7 @@ func TestEndpointSliceDiscoveryUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -640,8 +660,10 @@ func TestEndpointSliceDiscoveryEmptyEndpoints(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -652,6 +674,7 @@ func TestEndpointSliceDiscoveryEmptyEndpoints(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -662,6 +685,7 @@ func TestEndpointSliceDiscoveryEmptyEndpoints(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -674,6 +698,7 @@ func TestEndpointSliceDiscoveryEmptyEndpoints(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -724,8 +749,10 @@ func TestEndpointSliceDiscoveryWithService(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -736,6 +763,7 @@ func TestEndpointSliceDiscoveryWithService(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -746,6 +774,7 @@ func TestEndpointSliceDiscoveryWithService(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -758,6 +787,7 @@ func TestEndpointSliceDiscoveryWithService(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -824,8 +854,10 @@ func TestEndpointSliceDiscoveryWithServiceUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -836,6 +868,7 @@ func TestEndpointSliceDiscoveryWithServiceUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -846,6 +879,7 @@ func TestEndpointSliceDiscoveryWithServiceUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -858,6 +892,7 @@ func TestEndpointSliceDiscoveryWithServiceUpdate(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -914,8 +949,10 @@ func TestEndpointsSlicesDiscoveryWithNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -929,6 +966,7 @@ func TestEndpointsSlicesDiscoveryWithNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -939,6 +977,7 @@ func TestEndpointsSlicesDiscoveryWithNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -951,6 +990,7 @@ func TestEndpointsSlicesDiscoveryWithNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1014,8 +1054,10 @@ func TestEndpointsSlicesDiscoveryWithUpdatedNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -1029,6 +1071,7 @@ func TestEndpointsSlicesDiscoveryWithUpdatedNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1039,6 +1082,7 @@ func TestEndpointsSlicesDiscoveryWithUpdatedNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1051,6 +1095,7 @@ func TestEndpointsSlicesDiscoveryWithUpdatedNodeMetadata(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1160,8 +1205,10 @@ func TestEndpointSliceDiscoveryNamespaces(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -1172,6 +1219,7 @@ func TestEndpointSliceDiscoveryNamespaces(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -1182,6 +1230,7 @@ func TestEndpointSliceDiscoveryNamespaces(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -1194,6 +1243,7 @@ func TestEndpointSliceDiscoveryNamespaces(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1308,8 +1358,10 @@ func TestEndpointSliceDiscoveryOwnNamespace(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":        "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating":    "false",
 						"__meta_kubernetes_endpointslice_endpoint_hostname":                  "testendpoint1",
+						"__meta_kubernetes_endpointslice_endpoint_node_name":                 "foobar",
 						"__meta_kubernetes_endpointslice_endpoint_topology_present_topology": "true",
 						"__meta_kubernetes_endpointslice_endpoint_topology_topology":         "value",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                      "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                               "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":                  "http",
 						"__meta_kubernetes_endpointslice_port_name":                          "testport",
@@ -1320,6 +1372,7 @@ func TestEndpointSliceDiscoveryOwnNamespace(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1b",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -1330,6 +1383,7 @@ func TestEndpointSliceDiscoveryOwnNamespace(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "false",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "true",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1c",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
 						"__meta_kubernetes_endpointslice_port_protocol":                   "TCP",
@@ -1342,6 +1396,7 @@ func TestEndpointSliceDiscoveryOwnNamespace(t *testing.T) {
 						"__meta_kubernetes_endpointslice_endpoint_conditions_ready":       "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_serving":     "true",
 						"__meta_kubernetes_endpointslice_endpoint_conditions_terminating": "false",
+						"__meta_kubernetes_endpointslice_endpoint_zone":                   "us-east-1a",
 						"__meta_kubernetes_endpointslice_port":                            "9000",
 						"__meta_kubernetes_endpointslice_port_app_protocol":               "http",
 						"__meta_kubernetes_endpointslice_port_name":                       "testport",
@@ -1405,3 +1460,41 @@ func TestEndpointSliceDiscoveryEmptyPodStatus(t *testing.T) {
 		expectedRes:      map[string]*targetgroup.Group{},
 	}.Run(t)
 }
+
+// TestEndpointSliceInfIndexersCount makes sure that RoleEndpointSlice discovery
+// sets up indexing for the main Kube informer only when needed.
+// See: https://github.com/prometheus/prometheus/pull/13554#discussion_r1490965817
+func TestEndpointSliceInfIndexersCount(t *testing.T) {
+	tests := []struct {
+		name             string
+		withNodeMetadata bool
+	}{
+		{"with node metadata", true},
+		{"without node metadata", false},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			var (
+				n                    *Discovery
+				mainInfIndexersCount int
+			)
+			if tc.withNodeMetadata {
+				mainInfIndexersCount = 1
+				n, _ = makeDiscoveryWithMetadata(RoleEndpointSlice, NamespaceDiscovery{}, AttachMetadataConfig{Node: true})
+			} else {
+				n, _ = makeDiscovery(RoleEndpointSlice, NamespaceDiscovery{})
+			}
+
+			k8sDiscoveryTest{
+				discovery: n,
+				afterStart: func() {
+					n.RLock()
+					defer n.RUnlock()
+					require.Len(t, n.discoverers, 1)
+					require.Len(t, n.discoverers[0].(*EndpointSlice).endpointSliceInf.GetIndexer().GetIndexers(), mainInfIndexersCount)
+				},
+			}.Run(t)
+		})
+	}
+}

discovery/kubernetes/kubernetes.go

@@ -123,12 +123,17 @@ type SDConfig struct {
 	AttachMetadata     AttachMetadataConfig    `yaml:"attach_metadata,omitempty"`
 }
 
+// NewDiscovererMetrics implements discovery.Config.
+func (*SDConfig) NewDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	return newDiscovererMetrics(reg, rmi)
+}
+
 // Name returns the name of the Config.
 func (*SDConfig) Name() string { return "kubernetes" }
 
 // NewDiscoverer returns a Discoverer for the Config.
 func (c *SDConfig) NewDiscoverer(opts discovery.DiscovererOptions) (discovery.Discoverer, error) {
-	return New(opts.Logger, opts.Registerer, c)
+	return New(opts.Logger, opts.Metrics, c)
 }
 
 // SetDirectory joins any relative file paths with dir.
@@ -265,8 +270,7 @@ type Discovery struct {
 	selectors          roleSelector
 	ownNamespace       string
 	attachMetadata     AttachMetadataConfig
-	eventCount         *prometheus.CounterVec
-	metricRegisterer   discovery.MetricRegisterer
+	metrics            *kubernetesMetrics
 }
 
 func (d *Discovery) getNamespaces() []string {
@@ -285,7 +289,12 @@ func (d *Discovery) getNamespaces() []string {
 }
 
 // New creates a new Kubernetes discovery for the given role.
-func New(l log.Logger, reg prometheus.Registerer, conf *SDConfig) (*Discovery, error) {
+func New(l log.Logger, metrics discovery.DiscovererMetrics, conf *SDConfig) (*Discovery, error) {
+	m, ok := metrics.(*kubernetesMetrics)
+	if !ok {
+		return nil, fmt.Errorf("invalid discovery metrics type")
+	}
+
 	if l == nil {
 		l = log.NewNopLogger()
 	}
@@ -302,7 +311,7 @@ func New(l log.Logger, reg prometheus.Registerer, conf *SDConfig) (*Discovery, e
 		}
 	case conf.APIServer.URL == nil:
 		// Use the Kubernetes provided pod service account
-		// as described in https://kubernetes.io/docs/admin/service-accounts-admin/
+		// as described in https://kubernetes.io/docs/tasks/run-application/access-api-from-pod/#using-official-client-libraries
 		kcfg, err = rest.InClusterConfig()
 		if err != nil {
 			return nil, err
@@ -348,34 +357,7 @@ func New(l log.Logger, reg prometheus.Registerer, conf *SDConfig) (*Discovery, e
 		selectors:          mapSelector(conf.Selectors),
 		ownNamespace:       ownNamespace,
 		attachMetadata:     conf.AttachMetadata,
-		eventCount: prometheus.NewCounterVec(
-			prometheus.CounterOpts{
-				Namespace: discovery.KubernetesMetricsNamespace,
-				Name:      "events_total",
-				Help:      "The number of Kubernetes events handled.",
-			},
-			[]string{"role", "event"},
-		),
-	}
-
-	d.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{d.eventCount})
-
-	// Initialize metric vectors.
-	for _, role := range []string{
-		RoleEndpointSlice.String(),
-		RoleEndpoint.String(),
-		RoleNode.String(),
-		RolePod.String(),
-		RoleService.String(),
-		RoleIngress.String(),
-	} {
-		for _, evt := range []string{
-			MetricLabelRoleAdd,
-			MetricLabelRoleDelete,
-			MetricLabelRoleUpdate,
-		} {
-			d.eventCount.WithLabelValues(role, evt)
-		}
+		metrics:            m,
 	}
 
 	return d, nil
@@ -415,13 +397,6 @@ const resyncDisabled = 0
 func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 	d.Lock()
 
-	err := d.metricRegisterer.RegisterMetrics()
-	if err != nil {
-		level.Error(d.logger).Log("msg", "Unable to register metrics", "err", err.Error())
-		return
-	}
-	defer d.metricRegisterer.UnregisterMetrics()
-
 	namespaces := d.getNamespaces()
 
 	switch d.role {
@@ -510,10 +485,10 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 			eps := NewEndpointSlice(
 				log.With(d.logger, "role", "endpointslice"),
 				informer,
-				cache.NewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
-				cache.NewSharedInformer(plw, &apiv1.Pod{}, resyncDisabled),
+				d.mustNewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
+				d.mustNewSharedInformer(plw, &apiv1.Pod{}, resyncDisabled),
 				nodeInf,
-				d.eventCount,
+				d.metrics.eventCount,
 			)
 			d.discoverers = append(d.discoverers, eps)
 			go eps.endpointSliceInf.Run(ctx.Done())
@@ -570,10 +545,10 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 			eps := NewEndpoints(
 				log.With(d.logger, "role", "endpoint"),
 				d.newEndpointsByNodeInformer(elw),
-				cache.NewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
-				cache.NewSharedInformer(plw, &apiv1.Pod{}, resyncDisabled),
+				d.mustNewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
+				d.mustNewSharedInformer(plw, &apiv1.Pod{}, resyncDisabled),
 				nodeInf,
-				d.eventCount,
+				d.metrics.eventCount,
 			)
 			d.discoverers = append(d.discoverers, eps)
 			go eps.endpointsInf.Run(ctx.Done())
@@ -605,7 +580,7 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 				log.With(d.logger, "role", "pod"),
 				d.newPodsByNodeInformer(plw),
 				nodeInformer,
-				d.eventCount,
+				d.metrics.eventCount,
 			)
 			d.discoverers = append(d.discoverers, pod)
 			go pod.podInf.Run(ctx.Done())
@@ -627,8 +602,8 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 			}
 			svc := NewService(
 				log.With(d.logger, "role", "service"),
-				cache.NewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
-				d.eventCount,
+				d.mustNewSharedInformer(slw, &apiv1.Service{}, resyncDisabled),
+				d.metrics.eventCount,
 			)
 			d.discoverers = append(d.discoverers, svc)
 			go svc.informer.Run(ctx.Done())
@@ -666,7 +641,7 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 						return i.Watch(ctx, options)
 					},
 				}
-				informer = cache.NewSharedInformer(ilw, &networkv1.Ingress{}, resyncDisabled)
+				informer = d.mustNewSharedInformer(ilw, &networkv1.Ingress{}, resyncDisabled)
 			} else {
 				i := d.client.NetworkingV1beta1().Ingresses(namespace)
 				ilw := &cache.ListWatch{
@@ -681,19 +656,19 @@ func (d *Discovery) Run(ctx context.Context, ch chan<- []*targetgroup.Group) {
 						return i.Watch(ctx, options)
 					},
 				}
-				informer = cache.NewSharedInformer(ilw, &v1beta1.Ingress{}, resyncDisabled)
+				informer = d.mustNewSharedInformer(ilw, &v1beta1.Ingress{}, resyncDisabled)
 			}
 			ingress := NewIngress(
 				log.With(d.logger, "role", "ingress"),
 				informer,
-				d.eventCount,
+				d.metrics.eventCount,
 			)
 			d.discoverers = append(d.discoverers, ingress)
 			go ingress.informer.Run(ctx.Done())
 		}
 	case RoleNode:
 		nodeInformer := d.newNodeInformer(ctx)
-		node := NewNode(log.With(d.logger, "role", "node"), nodeInformer, d.eventCount)
+		node := NewNode(log.With(d.logger, "role", "node"), nodeInformer, d.metrics.eventCount)
 		d.discoverers = append(d.discoverers, node)
 		go node.informer.Run(ctx.Done())
 	default:
@@ -772,7 +747,7 @@ func (d *Discovery) newNodeInformer(ctx context.Context) cache.SharedInformer {
 			return d.client.CoreV1().Nodes().Watch(ctx, options)
 		},
 	}
-	return cache.NewSharedInformer(nlw, &apiv1.Node{}, resyncDisabled)
+	return d.mustNewSharedInformer(nlw, &apiv1.Node{}, resyncDisabled)
 }
 
 func (d *Discovery) newPodsByNodeInformer(plw *cache.ListWatch) cache.SharedIndexInformer {
@@ -787,13 +762,28 @@ func (d *Discovery) newPodsByNodeInformer(plw *cache.ListWatch) cache.SharedInde
 		}
 	}
 
-	return cache.NewSharedIndexInformer(plw, &apiv1.Pod{}, resyncDisabled, indexers)
+	return d.mustNewSharedIndexInformer(plw, &apiv1.Pod{}, resyncDisabled, indexers)
 }
 
 func (d *Discovery) newEndpointsByNodeInformer(plw *cache.ListWatch) cache.SharedIndexInformer {
 	indexers := make(map[string]cache.IndexFunc)
+	indexers[podIndex] = func(obj interface{}) ([]string, error) {
+		e, ok := obj.(*apiv1.Endpoints)
+		if !ok {
+			return nil, fmt.Errorf("object is not endpoints")
+		}
+		var pods []string
+		for _, target := range e.Subsets {
+			for _, addr := range target.Addresses {
+				if addr.TargetRef != nil && addr.TargetRef.Kind == "Pod" {
+					pods = append(pods, namespacedName(addr.TargetRef.Namespace, addr.TargetRef.Name))
+				}
+			}
+		}
+		return pods, nil
+	}
 	if !d.attachMetadata.Node {
-		return cache.NewSharedIndexInformer(plw, &apiv1.Endpoints{}, resyncDisabled, indexers)
+		return d.mustNewSharedIndexInformer(plw, &apiv1.Endpoints{}, resyncDisabled, indexers)
 	}
 
 	indexers[nodeIndex] = func(obj interface{}) ([]string, error) {
@@ -819,13 +809,13 @@ func (d *Discovery) newEndpointsByNodeInformer(plw *cache.ListWatch) cache.Share
 		return nodes, nil
 	}
 
-	return cache.NewSharedIndexInformer(plw, &apiv1.Endpoints{}, resyncDisabled, indexers)
+	return d.mustNewSharedIndexInformer(plw, &apiv1.Endpoints{}, resyncDisabled, indexers)
 }
 
 func (d *Discovery) newEndpointSlicesByNodeInformer(plw *cache.ListWatch, object runtime.Object) cache.SharedIndexInformer {
 	indexers := make(map[string]cache.IndexFunc)
 	if !d.attachMetadata.Node {
-		cache.NewSharedIndexInformer(plw, &disv1.EndpointSlice{}, resyncDisabled, indexers)
+		return d.mustNewSharedIndexInformer(plw, object, resyncDisabled, indexers)
 	}
 
 	indexers[nodeIndex] = func(obj interface{}) ([]string, error) {
@@ -864,7 +854,32 @@ func (d *Discovery) newEndpointSlicesByNodeInformer(plw *cache.ListWatch, object
 		return nodes, nil
 	}
 
-	return cache.NewSharedIndexInformer(plw, object, resyncDisabled, indexers)
+	return d.mustNewSharedIndexInformer(plw, object, resyncDisabled, indexers)
+}
+
+func (d *Discovery) informerWatchErrorHandler(r *cache.Reflector, err error) {
+	d.metrics.failuresCount.Inc()
+	cache.DefaultWatchErrorHandler(r, err)
+}
+
+func (d *Discovery) mustNewSharedInformer(lw cache.ListerWatcher, exampleObject runtime.Object, defaultEventHandlerResyncPeriod time.Duration) cache.SharedInformer {
+	informer := cache.NewSharedInformer(lw, exampleObject, defaultEventHandlerResyncPeriod)
+	// Invoking SetWatchErrorHandler should fail only if the informer has been started beforehand.
+	// Such a scenario would suggest an incorrect use of the API, thus the panic.
+	if err := informer.SetWatchErrorHandler(d.informerWatchErrorHandler); err != nil {
+		panic(err)
+	}
+	return informer
+}
+
+func (d *Discovery) mustNewSharedIndexInformer(lw cache.ListerWatcher, exampleObject runtime.Object, defaultEventHandlerResyncPeriod time.Duration, indexers cache.Indexers) cache.SharedIndexInformer {
+	informer := cache.NewSharedIndexInformer(lw, exampleObject, defaultEventHandlerResyncPeriod, indexers)
+	// Invoking SetWatchErrorHandler should fail only if the informer has been started beforehand.
+	// Such a scenario would suggest an incorrect use of the API, thus the panic.
+	if err := informer.SetWatchErrorHandler(d.informerWatchErrorHandler); err != nil {
+		panic(err)
+	}
+	return informer
 }
 
 func checkDiscoveryV1Supported(client kubernetes.Interface) (bool, error) {
@@ -897,3 +912,7 @@ func addObjectMetaLabels(labelSet model.LabelSet, objectMeta metav1.ObjectMeta,
 		labelSet[model.LabelName(metaLabelPrefix+string(role)+"_annotationpresent_"+ln)] = presentValue
 	}
 }
+
+func namespacedName(namespace, name string) string {
+	return namespace + "/" + name
+}

discovery/kubernetes/kubernetes_test.go

@@ -21,12 +21,16 @@ import (
 	"time"
 
 	"github.com/go-kit/log"
+	prom_testutil "github.com/prometheus/client_golang/prometheus/testutil"
 	"github.com/stretchr/testify/require"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/version"
+	"k8s.io/apimachinery/pkg/watch"
 	fakediscovery "k8s.io/client-go/discovery/fake"
 	"k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/kubernetes/fake"
+	kubetesting "k8s.io/client-go/testing"
 	"k8s.io/client-go/tools/cache"
 
 	"github.com/prometheus/client_golang/prometheus"
@@ -51,24 +55,29 @@ func makeDiscoveryWithVersion(role Role, nsDiscovery NamespaceDiscovery, k8sVer
 	fakeDiscovery, _ := clientset.Discovery().(*fakediscovery.FakeDiscovery)
 	fakeDiscovery.FakedServerVersion = &version.Info{GitVersion: k8sVer}
 
+	reg := prometheus.NewRegistry()
+	refreshMetrics := discovery.NewRefreshMetrics(reg)
+	metrics := newDiscovererMetrics(reg, refreshMetrics)
+	err := metrics.Register()
+	if err != nil {
+		panic(err)
+	}
+	// TODO(ptodev): Unregister the metrics at the end of the test.
+
+	kubeMetrics, ok := metrics.(*kubernetesMetrics)
+	if !ok {
+		panic("invalid discovery metrics type")
+	}
+
 	d := &Discovery{
 		client:             clientset,
 		logger:             log.NewNopLogger(),
 		role:               role,
 		namespaceDiscovery: &nsDiscovery,
 		ownNamespace:       "own-ns",
-		eventCount: prometheus.NewCounterVec(
-			prometheus.CounterOpts{
-				Namespace: discovery.KubernetesMetricsNamespace,
-				Name:      "events_total",
-				Help:      "The number of Kubernetes events handled.",
-			},
-			[]string{"role", "event"},
-		),
+		metrics:            kubeMetrics,
 	}
 
-	d.metricRegisterer = discovery.NewMetricRegisterer(prometheus.NewRegistry(), []prometheus.Collector{d.eventCount})
-
 	return d, clientset
 }
 
@@ -123,17 +132,11 @@ func (d k8sDiscoveryTest) Run(t *testing.T) {
 	}
 
 	resChan := make(chan map[string]*targetgroup.Group)
-	go readResultWithTimeout(t, ch, d.expectedMaxItems, time.Second, resChan)
+	go readResultWithTimeout(t, ctx, ch, d.expectedMaxItems, time.Second, resChan)
 
 	dd, ok := d.discovery.(hasSynced)
-	if !ok {
-		t.Errorf("discoverer does not implement hasSynced interface")
-		return
-	}
-	if !cache.WaitForCacheSync(ctx.Done(), dd.hasSynced) {
-		t.Errorf("discoverer failed to sync: %v", dd)
-		return
-	}
+	require.True(t, ok, "discoverer does not implement hasSynced interface")
+	require.True(t, cache.WaitForCacheSync(ctx.Done(), dd.hasSynced), "discoverer failed to sync: %v", dd)
 
 	if d.afterStart != nil {
 		d.afterStart()
@@ -142,13 +145,18 @@ func (d k8sDiscoveryTest) Run(t *testing.T) {
 	if d.expectedRes != nil {
 		res := <-resChan
 		requireTargetGroups(t, d.expectedRes, res)
+	} else {
+		// Stop readResultWithTimeout and wait for it.
+		cancel()
+		<-resChan
 	}
 }
 
 // readResultWithTimeout reads all targetgroups from channel with timeout.
 // It merges targetgroups by source and sends the result to result channel.
-func readResultWithTimeout(t *testing.T, ch <-chan []*targetgroup.Group, max int, timeout time.Duration, resChan chan<- map[string]*targetgroup.Group) {
+func readResultWithTimeout(t *testing.T, ctx context.Context, ch <-chan []*targetgroup.Group, max int, stopAfter time.Duration, resChan chan<- map[string]*targetgroup.Group) {
 	res := make(map[string]*targetgroup.Group)
+	timeout := time.After(stopAfter)
 Loop:
 	for {
 		select {
@@ -163,12 +171,15 @@ Loop:
 				// Reached max target groups we may get, break fast.
 				break Loop
 			}
-		case <-time.After(timeout):
+		case <-timeout:
 			// Because we use queue, an object that is created then
 			// deleted or updated may be processed only once.
 			// So possibly we may skip events, timed out here.
 			t.Logf("timed out, got %d (max: %d) items, some events are skipped", len(res), max)
 			break Loop
+		case <-ctx.Done():
+			t.Logf("stopped, got %d (max: %d) items", len(res), max)
+			break Loop
 		}
 	}
 
@@ -307,3 +318,39 @@ func TestCheckNetworkingV1Supported(t *testing.T) {
 		})
 	}
 }
+
+func TestFailuresCountMetric(t *testing.T) {
+	tests := []struct {
+		role             Role
+		minFailedWatches int
+	}{
+		{RoleNode, 1},
+		{RolePod, 1},
+		{RoleService, 1},
+		{RoleEndpoint, 3},
+		{RoleEndpointSlice, 3},
+		{RoleIngress, 1},
+	}
+
+	for _, tc := range tests {
+		tc := tc
+		t.Run(string(tc.role), func(t *testing.T) {
+			t.Parallel()
+
+			n, c := makeDiscovery(tc.role, NamespaceDiscovery{})
+			// The counter is initialized and no failures at the beginning.
+			require.Equal(t, float64(0), prom_testutil.ToFloat64(n.metrics.failuresCount))
+
+			// Simulate an error on watch requests.
+			c.Discovery().(*fakediscovery.FakeDiscovery).PrependWatchReactor("*", func(action kubetesting.Action) (bool, watch.Interface, error) {
+				return true, nil, apierrors.NewUnauthorized("unauthorized")
+			})
+
+			// Start the discovery.
+			k8sDiscoveryTest{discovery: n}.Run(t)
+
+			// At least the errors of the initial watches should be caught (watches are retried on errors).
+			require.GreaterOrEqual(t, prom_testutil.ToFloat64(n.metrics.failuresCount), float64(tc.minFailedWatches))
+		})
+	}
+}

discovery/kubernetes/metrics.go

@@ -0,0 +1,86 @@
+// Copyright 2015 The Prometheus Authors
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package kubernetes
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+
+	"github.com/prometheus/prometheus/discovery"
+)
+
+var _ discovery.DiscovererMetrics = (*kubernetesMetrics)(nil)
+
+type kubernetesMetrics struct {
+	eventCount    *prometheus.CounterVec
+	failuresCount prometheus.Counter
+
+	metricRegisterer discovery.MetricRegisterer
+}
+
+func newDiscovererMetrics(reg prometheus.Registerer, rmi discovery.RefreshMetricsInstantiator) discovery.DiscovererMetrics {
+	m := &kubernetesMetrics{
+		eventCount: prometheus.NewCounterVec(
+			prometheus.CounterOpts{
+				Namespace: discovery.KubernetesMetricsNamespace,
+				Name:      "events_total",
+				Help:      "The number of Kubernetes events handled.",
+			},
+			[]string{"role", "event"},
+		),
+		failuresCount: prometheus.NewCounter(
+			prometheus.CounterOpts{
+				Namespace: discovery.KubernetesMetricsNamespace,
+				Name:      "failures_total",
+				Help:      "The number of failed WATCH/LIST requests.",
+			},
+		),
+	}
+
+	m.metricRegisterer = discovery.NewMetricRegisterer(reg, []prometheus.Collector{
+		m.eventCount,
+		m.failuresCount,
+	})
+
+	// Initialize metric vectors.
+	for _, role := range []string{
+		RoleEndpointSlice.String(),
+		RoleEndpoint.String(),
+		RoleNode.String(),
+		RolePod.String(),
+		RoleService.String(),
+		RoleIngress.String(),
+	} {
+		for _, evt := range []string{
+			MetricLabelRoleAdd,
+			MetricLabelRoleDelete,
+			MetricLabelRoleUpdate,
+		} {
+			m.eventCount.WithLabelValues(role, evt)
+		}
+	}
+
+	m.failuresCount.Add(0)
+
+	return m
+}
+
+// Register implements discovery.DiscovererMetrics.
+func (m *kubernetesMetrics) Register() error {
+	return m.metricRegisterer.RegisterMetrics()
+}
+
+// Unregister implements discovery.DiscovererMetrics.
+func (m *kubernetesMetrics) Unregister() {
+	m.metricRegisterer.UnregisterMetrics()
+}

discovery/kubernetes/pod.go

@@ -33,7 +33,10 @@ import (
 	"github.com/prometheus/prometheus/discovery/targetgroup"
 )
 
-const nodeIndex = "node"
+const (
+	nodeIndex = "node"
+	podIndex  = "pod"
+)
 
 // Pod discovers new pod targets.
 type Pod struct {
@@ -326,7 +329,7 @@ func podSource(pod *apiv1.Pod) string {
 }
 
 func podSourceFromNamespaceAndName(namespace, name string) string {
-	return "pod/" + namespace + "/" + name
+	return "pod/" + namespacedName(namespace, name)
 }
 
 func podReady(pod *apiv1.Pod) model.LabelValue {

discovery/legacymanager/manager.go

@@ -42,7 +42,7 @@ type provider struct {
 }
 
 // NewManager is the Discovery Manager constructor.
-func NewManager(ctx context.Context, logger log.Logger, registerer prometheus.Registerer, options ...func(*Manager)) *Manager {
+func NewManager(ctx context.Context, logger log.Logger, registerer prometheus.Registerer, sdMetrics map[string]discovery.DiscovererMetrics, options ...func(*Manager)) *Manager {
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
@@ -55,6 +55,7 @@ func NewManager(ctx context.Context, logger log.Logger, registerer prometheus.Re
 		updatert:       5 * time.Second,
 		triggerSend:    make(chan struct{}, 1),
 		registerer:     registerer,
+		sdMetrics:      sdMetrics,
 	}
 	for _, option := range options {
 		option(mgr)
@@ -62,7 +63,7 @@ func NewManager(ctx context.Context, logger log.Logger, registerer prometheus.Re
 
 	// Register the metrics.
 	// We have to do this after setting all options, so that the name of the Manager is set.
-	if metrics, err := discovery.NewMetrics(registerer, mgr.name); err == nil {
+	if metrics, err := discovery.NewManagerMetrics(registerer, mgr.name); err == nil {
 		mgr.metrics = metrics
 	} else {
 		level.Error(logger).Log("msg", "Failed to create discovery manager metrics", "manager", mgr.name, "err", err)
@@ -108,7 +109,8 @@ type Manager struct {
 	// A registerer for all service discovery metrics.
 	registerer prometheus.Registerer
 
-	metrics *discovery.Metrics
+	metrics   *discovery.Metrics
+	sdMetrics map[string]discovery.DiscovererMetrics
 }
 
 // Run starts the background processing.
@@ -283,8 +285,8 @@ func (m *Manager) registerProviders(cfgs discovery.Configs, setName string) int
 		}
 		typ := cfg.Name()
 		d, err := cfg.NewDiscoverer(discovery.DiscovererOptions{
-			Logger:     log.With(m.logger, "discovery", typ, "config", setName),
-			Registerer: m.registerer,
+			Logger:  log.With(m.logger, "discovery", typ, "config", setName),
+			Metrics: m.sdMetrics[typ],
 		})
 		if err != nil {
 			level.Error(m.logger).Log("msg", "Cannot create service discovery", "err", err, "type", typ, "config", setName)