seaweedfs/.github/workflows/performance.yml

name: "Performance"

on:
  push:
    branches: [ master ]
    paths:
      - '**/*.go'
      - 'go.mod'
      - 'go.sum'
      - 'seaweed-volume/**'
      - 'seaweed-common/**'
      - 'test/perf/**'
      - '.github/workflows/performance.yml'
  workflow_dispatch:
    inputs:
      profile_duration:
        description: "CPU profiling duration in seconds"
        required: false
        default: "30"
        type: string
      benchmark_files:
        description: "Number of files for the throughput benchmark"
        required: false
        default: "100000"
        type: string
      benchmark_concurrency:
        description: "Concurrent read/write workers"
        required: false
        default: "16"
        type: string
      benchmark_size:
        description: "Simulated file size in bytes"
        required: false
        default: "1024"
        type: string
      s3_objects:
        description: "Number of objects for the S3 benchmark"
        required: false
        default: "20000"
        type: string
      s3_size:
        description: "S3 object size in bytes"
        required: false
        default: "4096"
        type: string

concurrency:
  group: ${{ github.head_ref || github.ref }}/performance
  cancel-in-progress: true

permissions:
  contents: read

env:
  VOL_SIZE_LIMIT: "1024"

jobs:
  performance-profile:
    name: CPU and Heap Profile
    runs-on: ubuntu-latest
    timeout-minutes: 30
    steps:
      - name: Check out code
        uses: actions/checkout@v7

      - name: Set up Go
        uses: actions/setup-go@v7
        with:
          go-version-file: 'go.mod'

      - name: Build weed
        run: go build -o weed_bin ./weed

      - name: Start server with profiling enabled
        run: |
          mkdir -p ./perfdata
          ./weed_bin -v=1 server -debug -debug.port=6060 -dir=./perfdata \
            -s3 -filer -volume.max=0 -master.volumeSizeLimitMB=100 \
            -s3.port=8000 -s3.config=./docker/compose/s3.json \
            > weed.log 2>&1 &
          echo "WEED_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 60); do
            if curl -sf http://localhost:9333/dir/status >/dev/null 2>&1; then
              echo "master is ready"
              break
            fi
            sleep 1
          done
          # give the volume server a moment to register with the master
          sleep 3

      - name: Start memory sampler
        run: |
          bash test/perf/mem_sample.sh mem-profile.csv "server=${WEED_PID}" &
          echo "SAMPLER_PID=$!" >> "$GITHUB_ENV"

      - name: Capture profiles under load
        run: |
          DURATION="${{ github.event.inputs.profile_duration || '30' }}"
          # drive write load so the sampled profile reflects real work
          ./weed_bin benchmark -master=localhost:9333 -writeOnly \
            -c=16 -n=5000000 -size=1024 > benchmark-load.log 2>&1 &
          echo "Sampling CPU profile for ${DURATION}s..."
          curl -s "http://localhost:6060/debug/pprof/profile?seconds=${DURATION}" -o cpu.pprof
          curl -s "http://localhost:6060/debug/pprof/heap" -o heap.pprof
          curl -s "http://localhost:6060/debug/pprof/goroutine?debug=1" -o goroutine.txt
          go tool pprof -top -nodecount=50 cpu.pprof > cpu-top.txt 2>/dev/null || true
          go tool pprof -top -nodecount=50 -sample_index=inuse_space heap.pprof > heap-top.txt 2>/dev/null || true

      - name: Record memory usage
        if: always()
        run: |
          kill -TERM "${SAMPLER_PID}" 2>/dev/null || true
          sleep 2
          {
            echo "## Memory usage (peak RSS)"
            echo '```'
            if [ -f mem-profile.csv.peak ]; then
              awk -F'\t' '{printf "%-10s %8d KB  (%.1f MB)\n", $1, $2, $2/1024}' mem-profile.csv.peak
            else
              echo "no memory samples captured"
            fi
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Profile summary
        if: always()
        run: |
          {
            echo "## CPU profile (top functions)"
            echo '```'
            head -45 cpu-top.txt 2>/dev/null || echo "no cpu profile captured"
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Stop server
        if: always()
        run: kill "${WEED_PID}" 2>/dev/null || true

      - name: Show server log on failure
        if: failure()
        run: tail -200 weed.log || true

      - name: Upload profiles
        if: always()
        uses: actions/upload-artifact@v7
        with:
          name: performance-profile-${{ github.run_number }}
          path: |
            cpu.pprof
            heap.pprof
            cpu-top.txt
            heap-top.txt
            goroutine.txt
            mem-profile.csv
            mem-profile.csv.peak
            weed.log
          retention-days: 30

  benchmark:
    name: Throughput Benchmark (${{ matrix.impl }} volume)
    runs-on: ubuntu-latest
    timeout-minutes: 60
    strategy:
      fail-fast: false
      matrix:
        impl: [go, rust]
    env:
      IMPL: ${{ matrix.impl }}
    steps:
      - name: Check out code
        uses: actions/checkout@v7

      - name: Set up Go
        uses: actions/setup-go@v7
        with:
          go-version-file: 'go.mod'

      - name: Install Rust toolchain
        if: matrix.impl == 'rust'
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo registry and target
        if: matrix.impl == 'rust'
        uses: actions/cache@v6
        with:
          path: |
            ~/.cargo/registry
            ~/.cargo/git
            seaweed-volume/target
          key: rust-${{ hashFiles('seaweed-volume/Cargo.lock') }}
          restore-keys: |
            rust-

      - name: Build weed
        run: go build -o weed_bin ./weed

      - name: Build Rust volume server
        if: matrix.impl == 'rust'
        run: cd seaweed-volume && cargo build --release

      - name: Start master and volume server
        run: |
          mkdir -p ./perfdata/master ./perfdata/vol
          ./weed_bin -v=1 master -ip=127.0.0.1 -port=9333 \
            -mdir=./perfdata/master -peers=none \
            -volumeSizeLimitMB="${VOL_SIZE_LIMIT}" -defaultReplication=000 \
            > master.log 2>&1 &
          echo "MASTER_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 60); do
            if curl -sf http://localhost:9333/dir/status >/dev/null 2>&1; then
              echo "master is ready"
              break
            fi
            sleep 1
          done
          if [ "${IMPL}" = "rust" ]; then
            ./seaweed-volume/target/release/weed-volume \
              --master 127.0.0.1:9333 --ip 127.0.0.1 --ip.bind 127.0.0.1 \
              --port 8080 --dir ./perfdata/vol --max 100 --preStopSeconds 0 \
              > volume.log 2>&1 &
          else
            ./weed_bin -v=1 volume -master=127.0.0.1:9333 -ip=127.0.0.1 \
              -port=8080 -dir=./perfdata/vol -max=100 \
              > volume.log 2>&1 &
          fi
          echo "VOLUME_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 60); do
            if curl -sf http://localhost:8080/status >/dev/null 2>&1; then
              echo "volume server is ready"
              break
            fi
            sleep 1
          done
          # let the volume server register with the master via heartbeat
          sleep 3

      - name: Start memory sampler
        run: |
          bash test/perf/mem_sample.sh mem-benchmark.csv \
            "master=${MASTER_PID}" "volume=${VOLUME_PID}" &
          echo "SAMPLER_PID=$!" >> "$GITHUB_ENV"

      - name: Run throughput benchmark
        run: |
          N="${{ github.event.inputs.benchmark_files || '100000' }}"
          C="${{ github.event.inputs.benchmark_concurrency || '16' }}"
          SIZE="${{ github.event.inputs.benchmark_size || '1024' }}"
          ./weed_bin benchmark -master=localhost:9333 \
            -c="${C}" -n="${N}" -size="${SIZE}" 2>&1 | tee benchmark-results.txt

      - name: Run Go micro-benchmarks
        if: matrix.impl == 'go'
        continue-on-error: true
        run: |
          go test -run='^$' -bench=. -benchmem -benchtime=10x \
            ./weed/topology/... ./weed/util/log_buffer/... ./weed/util/buffered_queue/... \
            2>&1 | tee go-benchmarks.txt

      - name: Record memory usage
        if: always()
        run: |
          kill -TERM "${SAMPLER_PID}" 2>/dev/null || true
          sleep 2
          {
            echo "## Memory usage (peak RSS, ${IMPL} volume)"
            echo '```'
            if [ -f mem-benchmark.csv.peak ]; then
              awk -F'\t' '{printf "%-10s %8d KB  (%.1f MB)\n", $1, $2, $2/1024}' mem-benchmark.csv.peak
            else
              echo "no memory samples captured"
            fi
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Benchmark summary
        if: always()
        run: |
          {
            echo "## Throughput benchmark (${IMPL} volume)"
            echo '```'
            grep -E "Concurrency Level|Time taken|Completed requests|Failed requests|Requests per second|Transfer rate" \
              benchmark-results.txt 2>/dev/null || echo "no benchmark results captured"
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Stop processes
        if: always()
        run: |
          kill "${VOLUME_PID}" "${MASTER_PID}" 2>/dev/null || true

      - name: Show logs on failure
        if: failure()
        run: |
          echo "=== master.log ==="; tail -100 master.log 2>/dev/null || true
          echo "=== volume.log ==="; tail -200 volume.log 2>/dev/null || true

      - name: Upload benchmark results
        if: always()
        uses: actions/upload-artifact@v7
        with:
          name: benchmark-results-${{ matrix.impl }}-${{ github.run_number }}
          path: |
            benchmark-results.txt
            go-benchmarks.txt
            mem-benchmark.csv
            mem-benchmark.csv.peak
            master.log
            volume.log
          retention-days: 7

  s3-benchmark:
    name: S3 Read/Write Benchmark (${{ matrix.impl }} volume)
    runs-on: ubuntu-latest
    timeout-minutes: 60
    strategy:
      fail-fast: false
      matrix:
        impl: [go, rust]
    env:
      IMPL: ${{ matrix.impl }}
    steps:
      - name: Check out code
        uses: actions/checkout@v7

      - name: Set up Go
        uses: actions/setup-go@v7
        with:
          go-version-file: 'go.mod'

      - name: Install Rust toolchain
        if: matrix.impl == 'rust'
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo registry and target
        if: matrix.impl == 'rust'
        uses: actions/cache@v6
        with:
          path: |
            ~/.cargo/registry
            ~/.cargo/git
            seaweed-volume/target
          key: rust-${{ hashFiles('seaweed-volume/Cargo.lock') }}
          restore-keys: |
            rust-

      - name: Build weed and S3 load tool
        run: |
          go build -o weed_bin ./weed
          go build -o s3bench ./test/s3/benchmark

      - name: Build Rust volume server
        if: matrix.impl == 'rust'
        run: cd seaweed-volume && cargo build --release

      - name: Start cluster with S3 gateway
        run: |
          mkdir -p ./perfdata/master ./perfdata/vol ./perfdata/filer
          ./weed_bin -v=1 master -ip=127.0.0.1 -port=9333 \
            -mdir=./perfdata/master -peers=none \
            -volumeSizeLimitMB="${VOL_SIZE_LIMIT}" -defaultReplication=000 \
            > master.log 2>&1 &
          echo "MASTER_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 60); do
            if curl -sf http://localhost:9333/dir/status >/dev/null 2>&1; then
              echo "master is ready"
              break
            fi
            sleep 1
          done
          if [ "${IMPL}" = "rust" ]; then
            ./seaweed-volume/target/release/weed-volume \
              --master 127.0.0.1:9333 --ip 127.0.0.1 --ip.bind 127.0.0.1 \
              --port 8080 --dir ./perfdata/vol --max 100 --preStopSeconds 0 \
              > volume.log 2>&1 &
          else
            ./weed_bin -v=1 volume -master=127.0.0.1:9333 -ip=127.0.0.1 \
              -port=8080 -dir=./perfdata/vol -max=100 \
              > volume.log 2>&1 &
          fi
          echo "VOLUME_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 60); do
            if curl -sf http://localhost:8080/status >/dev/null 2>&1; then
              echo "volume server is ready"
              break
            fi
            sleep 1
          done
          sleep 3
          ./weed_bin -v=1 filer -master=127.0.0.1:9333 -ip=127.0.0.1 -port=8888 \
            -s3 -s3.port=8000 -s3.config=./docker/compose/s3.json \
            > filer.log 2>&1 &
          echo "FILER_PID=$!" >> "$GITHUB_ENV"
          for i in $(seq 1 30); do
            if nc -z localhost 8000 2>/dev/null; then
              echo "s3 gateway is ready"
              break
            fi
            sleep 1
          done
          sleep 2

      - name: Start memory sampler
        run: |
          bash test/perf/mem_sample.sh mem-s3.csv \
            "master=${MASTER_PID}" "volume=${VOLUME_PID}" "filer=${FILER_PID}" &
          echo "SAMPLER_PID=$!" >> "$GITHUB_ENV"

      - name: Run S3 read/write benchmark
        run: |
          OBJECTS="${{ github.event.inputs.s3_objects || '20000' }}"
          C="${{ github.event.inputs.benchmark_concurrency || '16' }}"
          SIZE="${{ github.event.inputs.s3_size || '4096' }}"
          ./s3bench -endpoint=http://localhost:8000 \
            -access-key=some_access_key1 -secret-key=some_secret_key1 \
            -objects="${OBJECTS}" -size="${SIZE}" -concurrency="${C}" -mode=both \
            2>&1 | tee s3-benchmark-results.txt

      - name: Record memory usage
        if: always()
        run: |
          kill -TERM "${SAMPLER_PID}" 2>/dev/null || true
          sleep 2
          {
            echo "## Memory usage (peak RSS, ${IMPL} volume)"
            echo '```'
            if [ -f mem-s3.csv.peak ]; then
              awk -F'\t' '{printf "%-10s %8d KB  (%.1f MB)\n", $1, $2, $2/1024}' mem-s3.csv.peak
            else
              echo "no memory samples captured"
            fi
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: S3 benchmark summary
        if: always()
        run: |
          {
            echo "## S3 read/write benchmark (${IMPL} volume)"
            echo '```'
            grep -E "results:|Concurrency Level|Time taken|Completed requests|Failed requests|Requests per second|Transfer rate|Latency" \
              s3-benchmark-results.txt 2>/dev/null || echo "no S3 benchmark results captured"
            echo '```'
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Stop processes
        if: always()
        run: |
          kill "${FILER_PID}" "${VOLUME_PID}" "${MASTER_PID}" 2>/dev/null || true

      - name: Show logs on failure
        if: failure()
        run: |
          echo "=== master.log ==="; tail -100 master.log 2>/dev/null || true
          echo "=== volume.log ==="; tail -200 volume.log 2>/dev/null || true
          echo "=== filer.log ==="; tail -200 filer.log 2>/dev/null || true

      - name: Upload S3 benchmark results
        if: always()
        uses: actions/upload-artifact@v7
        with:
          name: s3-benchmark-results-${{ matrix.impl }}-${{ github.run_number }}
          path: |
            s3-benchmark-results.txt
            mem-s3.csv
            mem-s3.csv.peak
            master.log
            volume.log
            filer.log
          retention-days: 7