name: VMBench Long Range
on:
  workflow_dispatch:
    inputs:
      run_greptimedb:
        description: Run GreptimeDB
        type: boolean
        default: true
        required: false
      run_victoriametrics:
        description: Run VictoriaMetrics
        type: boolean
        default: false
        required: false
      range_preset:
        description: Long-range query window (fixed dataset per preset)
        type: choice
        options:
        - 1w
        - 1mo
        default: 1w
        required: true
      compare_greptimedb:
        description: Compare baseline and candidate sequentially on the same host (adds one full GreptimeDB run)
        type: boolean
        default: false
        required: false
      baseline_greptimedb_tag:
        description: Baseline Docker Hub greptime/greptimedb tag; required when comparison is enabled
        type: string
        default: v1.2.1
        required: false
      greptimedb_tag:
        description: Candidate Docker Hub greptime/greptimedb tag
        type: string
        default: latest
        required: true
      victoriametrics_tag:
        description: Docker Hub victoriametrics/victoria-metrics tag or tag@sha256 digest
        type: string
        default: v1.114.0@sha256:227ad53fd57a4b4430624898a1be0cac39431e1235fa0dc92f567cce38969344
        required: true
      runtime_image:
        description: 'Runtime image: auto = AWS ECR Public or Aliyun ACR by provider; explicit tag/digest overrides'
        type: string
        default: auto
        required: true
      provider:
        description: Cloud provider for the benchmark runner
        type: choice
        options: [Aliyun, AWS]
        default: Aliyun
        required: true
      instance_type:
        description: 'Instance type; auto = Aliyun: ecs.c9i.2xlarge, AWS: c7i.2xlarge (both 8 vCPU / 16 GiB). Enter a specific type to override.'
        type: string
        default: auto
        required: true
      benchmark_root:
        description: Absolute disk-backed work directory (outside the o11ybench checkout)
        type: string
        default: /home/runner/benchmark-data
        required: true
      system_disk_gib:
        description: 'System disk GiB: auto (or 0) = 80; enter 20..2048 to override'
        type: string
        default: auto
        required: true
      benchmark_timeout_minutes:
        description: Benchmark job timeout in minutes (1..360)
        type: number
        default: 360
        required: true
      janitor_ttl_hours:
        description: Instance lifetime before janitor eligibility (1..168 hours; at least timeout + 2h)
        type: number
        default: 8
        required: true
      db_cpus:
        description: CPU limit per DB container
        type: string
        default: '8'
        required: true
      db_memory:
        description: Memory limit per DB container
        type: string
        default: 16g
        required: true
permissions:
  contents: read
env:
  COMPARE_GREPTIMEDB: ${{ inputs.compare_greptimedb }}
  BASELINE_GREPTIMEDB_TAG: ${{ inputs.baseline_greptimedb_tag }}
  BENCHMARK_ROOT: ${{ inputs.benchmark_root }}
  RANGE_PRESET: ${{ inputs.range_preset }}
  GREPTIMEDB_TAG: ${{ inputs.greptimedb_tag }}
  VICTORIAMETRICS_TAG: ${{ inputs.victoriametrics_tag }}
  RUNTIME_IMAGE: ${{ inputs.runtime_image != 'auto' && inputs.runtime_image || inputs.provider == 'AWS' && 'public.ecr.aws/i8k6a5e1/greptime/o11ybenchruntime:b43785b56f42@sha256:e38811eb68caf3682ba56f5591a3e8e27e78e7983a923abb29fa2d7cd8fdbd46' || 'greptime-registry.cn-hangzhou.cr.aliyuncs.com/tools/o11ybench-runtime:b43785b56f42@sha256:e38811eb68caf3682ba56f5591a3e8e27e78e7983a923abb29fa2d7cd8fdbd46' }}
  PROVIDER: ${{ inputs.provider }}
  REQUESTED_INSTANCE_TYPE: ${{ inputs.instance_type }}
  SYSTEM_DISK_GIB: ${{ inputs.system_disk_gib }}
  BENCHMARK_TIMEOUT_MINUTES: ${{ inputs.benchmark_timeout_minutes }}
  JANITOR_TTL_HOURS: ${{ inputs.janitor_ttl_hours }}
  DB_CPUS: ${{ inputs.db_cpus }}
  DB_MEMORY: ${{ inputs.db_memory }}
jobs:
  validate:
    runs-on: ubuntu-latest
    timeout-minutes: 15
    outputs:
      instance_type: ${{ steps.runner.outputs.instance_type }}
      system_disk_gib: ${{ steps.plan.outputs.system_disk_gib }}
      runtime_image: ${{ steps.plan.outputs.runtime_image }}
      o11ybench_ref: ${{ steps.plan.outputs.o11ybench_ref }}
      run_id: ${{ steps.plan.outputs.run_id }}
      plan_artifact: ${{ steps.plan.outputs.plan_artifact }}
    steps:
    - uses: actions/checkout@v4
      with:
        path: greptimedb
        persist-credentials: false
    - name: Validate version comparison
      env:
        RUN_GREPTIMEDB: ${{ inputs.run_greptimedb }}
      run: |
        python3 greptimedb/tests/perf/test_benchmark_version_comparison.py
        python3 greptimedb/.github/scripts/benchmark-version-comparison.py validate
    - name: Resolve CI runner
      id: runner
      run: python3 greptimedb/.github/scripts/ci-runner-config.py --provider "$PROVIDER" --instance-type "$REQUESTED_INSTANCE_TYPE"
    - name: Test CI runner helpers
      run: |
        python3 greptimedb/tests/perf/test_aliyun_ecs_runner_scripts.py
        python3 greptimedb/tests/perf/test_ci_runner_scripts.py
    - name: Validate budgets and build warm plan
      id: plan
      env:
        RUN_GREPTIMEDB: ${{ inputs.run_greptimedb }}
        RUN_VICTORIAMETRICS: ${{ inputs.run_victoriametrics }}
      run: |
        set -euo pipefail
        [[ "$BENCHMARK_ROOT" == /* && "$BENCHMARK_ROOT" != / && "$BENCHMARK_ROOT" != */ ]]
        [[ "$BENCHMARK_ROOT" != /tmp && "$BENCHMARK_ROOT" != /tmp/* ]]
        [[ "$BENCHMARK_ROOT" != *$'\n'* && "$BENCHMARK_ROOT" != *$'\r'* ]]
        [[ "/${BENCHMARK_ROOT#/}/" != */../* && "/${BENCHMARK_ROOT#/}/" != */./* ]]
        if [[ "$SYSTEM_DISK_GIB" == auto || "$SYSTEM_DISK_GIB" == 0 ]]; then
          SYSTEM_DISK_GIB=80
        fi
        for value in "$SYSTEM_DISK_GIB" "$BENCHMARK_TIMEOUT_MINUTES" "$JANITOR_TTL_HOURS"; do
          [[ "$value" =~ ^[1-9][0-9]{0,3}$ ]] || { echo 'Resource budgets must be positive integers'; exit 1; }
        done
        ((SYSTEM_DISK_GIB >= 20 && SYSTEM_DISK_GIB <= 2048))
        ((BENCHMARK_TIMEOUT_MINUTES <= 360 && JANITOR_TTL_HOURS <= 168))
        # Match the existing ECS workflow's provision, teardown and queue headroom.
        ((JANITOR_TTL_HOURS * 60 >= BENCHMARK_TIMEOUT_MINUTES + 120))
        [[ "$RUNTIME_IMAGE" == *.cr.aliyuncs.com/* || "$RUNTIME_IMAGE" == public.ecr.aws/* ]]
        bash greptimedb/.github/scripts/pull-benchmark-image.sh "$RUNTIME_IMAGE"
        metadata=$(docker image inspect "$RUNTIME_IMAGE")
        resolved=$(jq -er '.[0].RepoDigests[0]' <<< "$metadata")
        revision=$(jq -er '.[0].Config.Labels["org.opencontainers.image.revision"]' <<< "$metadata")
        [[ "$revision" =~ ^[0-9a-f]{40}$ ]]
        printf 'runtime_image=%s\no11ybench_ref=%s\nsystem_disk_gib=%s\n' "$resolved" "$revision" "$SYSTEM_DISK_GIB" >> "$GITHUB_OUTPUT"
        printf 'run_id=long-range-%s-%s\nplan_artifact=vmbench-plan-%s-%s\n' "$GITHUB_RUN_ID" "$GITHUB_RUN_ATTEMPT" "$GITHUB_RUN_ID" "$GITHUB_RUN_ATTEMPT" >> "$GITHUB_OUTPUT"
        args=(--workload vmbench-long-range --range-preset "$RANGE_PRESET"
          --db-cpus "$DB_CPUS" --db-memory "$DB_MEMORY"
          --run-classification controlled_server_benchmark --no-open-browser)
        if [[ "$COMPARE_GREPTIMEDB" == true ]]; then
          args+=(--target "greptimedb:name=greptimedb-baseline,image=greptime/greptimedb:$BASELINE_GREPTIMEDB_TAG,url=http://127.0.0.1:14000/v1/prometheus/api/v1/query_range")
        fi
        # Tags become target-spec fields; reject separators before passing them to the planner.
        for target in greptimedb victoriametrics; do
          case "$target" in
            greptimedb) enabled=$RUN_GREPTIMEDB; tag=$GREPTIMEDB_TAG; repo=greptime/greptimedb; url=http://127.0.0.1:14000/v1/prometheus/api/v1/query_range ;;
            victoriametrics) enabled=$RUN_VICTORIAMETRICS; tag=$VICTORIAMETRICS_TAG; repo=victoriametrics/victoria-metrics; url=http://127.0.0.1:18428/api/v1/query_range ;;
          esac
          [[ "$enabled" == true ]] || continue
          [[ "$tag" =~ ^[a-zA-Z0-9_][a-zA-Z0-9_.-]{0,127}(@sha256:[0-9a-f]{64})?$ ]]
          args+=(--target "$target:name=$target,image=$repo:$tag,url=$url")
        done
        # The packaged planner owns profile, target and resource-limit validation.
        mkdir -p "$RUNNER_TEMP/vmbench-plans"
        # Keep the original warm defaults: serial 5/1 and c1/c4 30s/5s.
        docker run --rm "$resolved" scripts/o11ybench_plan.py "${args[@]}" \
          --run-id "long-range-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT-warm" \
          --output-root "$BENCHMARK_ROOT/o11ybench/o11ybench-long-range-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT-warm" \
          --cache-policy warm > "$RUNNER_TEMP/vmbench-plans/warm.json"
    - uses: actions/upload-artifact@v4
      with:
        name: vmbench-plan-${{ github.run_id }}-${{ github.run_attempt }}
        path: ${{ runner.temp }}/vmbench-plans/
        if-no-files-found: error
        retention-days: 1
  provision:
    needs: validate
    runs-on: ubuntu-latest
    timeout-minutes: 45
    outputs:
      label: ${{ steps.ecs.outputs.label || steps.aws.outputs.label }}
      instance_id: ${{ steps.ecs.outputs.instance_id || steps.aws.outputs.instance_id }}
      runner_name: ${{ steps.ecs.outputs.runner_name || steps.aws.outputs.runner_name }}
    steps:
    - uses: actions/checkout@v4
      with:
        path: greptimedb
        persist-credentials: false
    - name: Create ephemeral ECS runner
      if: inputs.provider == 'Aliyun'
      id: ecs
      uses: ./greptimedb/.github/actions/aliyun-ecs-create
      with:
        access-key-id: ${{ secrets.ALICLOUD_ECS_ACCESS_KEY_ID }}
        access-key-secret: ${{ secrets.ALICLOUD_ECS_ACCESS_KEY_SECRET }}
        github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
        region-id: ${{ vars.ALIYUN_ECS_REGION_ID }}
        vswitch-id: ${{ vars.ALIYUN_ECS_VSWITCH_ID }}
        security-group-id: ${{ vars.ALIYUN_ECS_SECURITY_GROUP_ID }}
        image-id: ${{ vars.QUERY_REGRESSION_ECS_IMAGE_ID }}
        instance-type: ${{ needs.validate.outputs.instance_type }}
        enable-docker: 'true'
        system-disk-gib: ${{ needs.validate.outputs.system_disk_gib }}
        ttl-hours: ${{ inputs.janitor_ttl_hours }}
        resource-group-id: ${{ vars.ALIYUN_ECS_RESOURCE_GROUP_ID }}
        runner-uid: ${{ vars.QUERY_REGRESSION_RUNNER_UID || '1001' }}
        runner-gid: ${{ vars.QUERY_REGRESSION_RUNNER_GID || '1001' }}
    - uses: aws-actions/configure-aws-credentials@v4
      if: inputs.provider == 'AWS'
      with:
        aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
        aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
        aws-region: ${{ vars.EC2_RUNNER_REGION }}
    - uses: astral-sh/setup-uv@v6
      if: inputs.provider == 'AWS'
    - name: Create ephemeral EC2 runner
      if: inputs.provider == 'AWS'
      id: aws
      env:
        GH_PERSONAL_ACCESS_TOKEN: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
        AWS_REGION: ${{ vars.EC2_RUNNER_REGION }}
        AWS_EC2_IMAGE_ID: ${{ vars.BENCHMARK_EC2_IMAGE_ID || 'auto' }}
        AWS_EC2_INSTANCE_TYPE: ${{ needs.validate.outputs.instance_type }}
        AWS_EC2_SUBNET_ID: ${{ vars.EC2_RUNNER_SUBNET_ID }}
        AWS_EC2_SECURITY_GROUP_ID: ${{ vars.EC2_RUNNER_SECURITY_GROUP_ID }}
        AWS_EC2_SYSTEM_DISK_GIB: ${{ needs.validate.outputs.system_disk_gib }}
        AWS_EC2_TTL_HOURS: ${{ inputs.janitor_ttl_hours }}
        AWS_EC2_RUN_ID: ${{ github.run_id }}-${{ github.run_attempt }}
      run: uv run greptimedb/.github/scripts/aws-ec2-runner-provision.py
  benchmark:
    needs:
    - validate
    - provision
    runs-on: ${{ needs.provision.outputs.label }}
    timeout-minutes: ${{ fromJSON(format('{0}', inputs.benchmark_timeout_minutes)) }}
    env:
      RESOLVED_RUNTIME_IMAGE: ${{ needs.validate.outputs.runtime_image }}
      RESOLVED_O11YBENCH_REF: ${{ needs.validate.outputs.o11ybench_ref }}
      RUN_ID: ${{ needs.validate.outputs.run_id }}
      ARTIFACT_ROOT: ${{ inputs.benchmark_root }}/o11ybench/o11ybench-${{ needs.validate.outputs.run_id }}
      METADATA_ROOT: ${{ github.workspace }}/vmbench-metadata
      O11YBENCH_DIR: ${{ github.workspace }}/o11ybench
      TSDG_PARALLELISM: '4'
      TSDG_BATCH_SIZE: '100000'
      O11YBENCH_GENERATOR_CPUS: '4'
    steps:
    - uses: actions/checkout@v4
      with:
        path: greptimedb
        persist-credentials: false
    - name: Set up benchmark dependencies
      uses: ./greptimedb/.github/actions/setup-benchmark
    - name: Checkout matching benchmark tools
      uses: actions/checkout@v4
      with:
        repository: GreptimeTeam/o11ybench
        token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
        ref: ${{ env.RESOLVED_O11YBENCH_REF }}
        path: o11ybench
        persist-credentials: false
    - uses: actions/download-artifact@v4
      with:
        name: ${{ needs.validate.outputs.plan_artifact }}
        path: ${{ env.METADATA_ROOT }}
    - name: Check runner and resolve database images
      run: |
        set -euo pipefail
        [[ "$(uname -m)" == x86_64 ]]
        test "$(git -C "$O11YBENCH_DIR" rev-parse HEAD)" = "$RESOLVED_O11YBENCH_REF"
        jq --version
        docker info
        mkdir -p "$BENCHMARK_ROOT/o11ybench"
        test "$(stat -f -c %T "$BENCHMARK_ROOT")" != tmpfs || { echo 'Benchmark root must be disk-backed'; exit 1; }
        df -h "$BENCHMARK_ROOT" /var/lib/docker
        # Fail before generation if either dataset space or Docker storage is too small.
        for path in "$BENCHMARK_ROOT" /var/lib/docker; do
          available=$(df -Pk "$path" | awk 'NR == 2 {print $4}')
          ((available >= 10 * 1024 * 1024)) || { echo "Need at least 10 GiB free on $path"; exit 1; }
        done
        bash greptimedb/.github/scripts/pull-benchmark-image.sh "$RESOLVED_RUNTIME_IMAGE"
        docker image inspect "$RESOLVED_RUNTIME_IMAGE" > "$METADATA_ROOT/runtime-image.json"
        test "$(jq -r '.[0].Config.Labels["org.opencontainers.image.revision"]' "$METADATA_ROOT/runtime-image.json")" = "$RESOLVED_O11YBENCH_REF"
        plan="$METADATA_ROOT/warm.json"
        while IFS=$'\t' read -r target reference; do
          echo "Resolving $target: $reference"
          bash greptimedb/.github/scripts/pull-benchmark-image.sh "$reference"
          docker image inspect "$reference" > "$METADATA_ROOT/$target-image.json"
          digest=$(jq -er '.[0].RepoDigests[0]' "$METADATA_ROOT/$target-image.json")
          jq --arg target "$target" --arg image "$digest" '.targets |= map(if .name == $target then .image = $image else . end)' "$plan" > "$plan.tmp"
          mv "$plan.tmp" "$plan"
        done < <(jq -r '.targets[] | [.name, .image] | @tsv' "$plan")
    - name: Review execution plan
      run: |
        set -euo pipefail
        python3 o11ybench/scripts/run_benchmark_plan.py \
          --plan "$METADATA_ROOT/warm.json" --runtime-image "$RESOLVED_RUNTIME_IMAGE" --dry-run \
          > "$METADATA_ROOT/warm-execution-plan.json"
        cat "$METADATA_ROOT/warm-execution-plan.json"
    - name: Warm serial and concurrent benchmarks
      run: |
        set -euo pipefail
        python3 -u o11ybench/scripts/run_benchmark_plan.py \
          --plan "$METADATA_ROOT/warm.json" --runtime-image "$RESOLVED_RUNTIME_IMAGE" --execute \
          2>&1 | tee "$METADATA_ROOT/warm-execution.log"
        python3 o11ybench/scripts/check_run_benchmark_plan.py --artifacts "$ARTIFACT_ROOT-warm"
    - name: Clean up owned containers
      if: always()
      run: |
        docker ps -aq --filter "label=o11ybench.run-id=$RUN_ID-warm" | xargs -r docker rm -fv
    - name: GreptimeDB version comparison
      if: always() && inputs.compare_greptimedb
      run: |
        python3 greptimedb/.github/scripts/benchmark-version-comparison.py report --kind long-range --root "$ARTIFACT_ROOT-warm" >> "$GITHUB_STEP_SUMMARY"
    - name: Stage benchmark artifacts
      if: always()
      run: |
        mkdir -p "$RUNNER_TEMP/vmbench-artifacts"
        if [[ -d "$METADATA_ROOT" ]]; then
          rsync -a --no-owner --no-group "$METADATA_ROOT" "$RUNNER_TEMP/vmbench-artifacts/"
        fi
        if [[ -d "$ARTIFACT_ROOT-warm" ]]; then
          # Exclude DB directories before traversal; they may be unreadable by the runner.
          rsync -a --no-owner --no-group --exclude='db-data/' --exclude='/data/*.bin' \
            "$ARTIFACT_ROOT-warm/" "$RUNNER_TEMP/vmbench-artifacts/$(basename "$ARTIFACT_ROOT-warm")/"
        fi
    - name: Upload HTML and results (no dataset or DB files)
      if: always()
      uses: actions/upload-artifact@v4
      with:
        name: vmbench-long-range-${{ github.run_id }}-${{ github.run_attempt }}
        path: ${{ runner.temp }}/vmbench-artifacts
        if-no-files-found: warn
        retention-days: 14
  teardown:
    needs:
    - provision
    - benchmark
    if: ${{ always() && (needs.provision.outputs.instance_id != '' || (inputs.provider == 'AWS' && needs.provision.result != 'skipped')) }}
    runs-on: ubuntu-latest
    timeout-minutes: 15
    steps:
    - uses: actions/checkout@v4
      with:
        path: greptimedb
        persist-credentials: false
    - name: Delete ECS and unregister runner
      if: inputs.provider == 'Aliyun'
      uses: ./greptimedb/.github/actions/aliyun-ecs-delete
      with:
        access-key-id: ${{ secrets.ALICLOUD_ECS_ACCESS_KEY_ID }}
        access-key-secret: ${{ secrets.ALICLOUD_ECS_ACCESS_KEY_SECRET }}
        github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
        region-id: ${{ vars.ALIYUN_ECS_REGION_ID }}
        instance-id: ${{ needs.provision.outputs.instance_id }}
        runner-name: ${{ needs.provision.outputs.runner_name }}
    - uses: aws-actions/configure-aws-credentials@v4
      if: inputs.provider == 'AWS'
      with:
        aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
        aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
        aws-region: ${{ vars.EC2_RUNNER_REGION }}
    - uses: astral-sh/setup-uv@v6
      if: inputs.provider == 'AWS'
    - name: Delete EC2 and unregister runner
      if: inputs.provider == 'AWS'
      env:
        GH_PERSONAL_ACCESS_TOKEN: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
        AWS_REGION: ${{ vars.EC2_RUNNER_REGION }}
        AWS_EC2_INSTANCE_ID: ${{ needs.provision.outputs.instance_id }}
        AWS_EC2_RUNNER_NAME: ${{ needs.provision.outputs.runner_name }}
        AWS_EC2_RUN_ID: ${{ github.run_id }}-${{ github.run_attempt }}
      run: uv run greptimedb/.github/scripts/aws-ec2-runner-teardown.py