[GHA testing] Add S3 express benchmark

.github/workflows/bench_main.yml

@@ -14,3 +14,8 @@ jobs:
     uses: ./.github/workflows/bench.yml
     with:
       ref: ${{ github.event.after }}
+  s3express-integration:
+    name: Benchmarks (s3express)
+    uses: ./.github/workflows/bench_s3express.yml
+    with:
+      ref: ${{ github.event.after }}

.github/workflows/bench_pr.yml

@@ -17,3 +17,10 @@ jobs:
     with:
       environment: PR benchmarks
       ref: ${{ github.event.pull_request.head.sha }}
+  s3express-integration:
+    name: Benchmarks (s3express)
+    uses: ./.github/workflows/bench_s3express.yml
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'performance') }}
+    with:
+      environment: PR benchmarks
+      ref: ${{ github.event.pull_request.head.sha }}

.github/workflows/bench_s3express.yml

@@ -0,0 +1,163 @@
+name: Benchmark (s3-express one zone)
+
+# We use environments to require approval to run benchmarks on PRs, but not on pushes to `main`
+# (which have been approved already since PRs are required for `main`).
+on:
+  workflow_call:
+    inputs:
+      environment:
+        type: string
+      ref:
+        required: true
+        type: string
+
+env:
+  RUST_BACKTRACE: 1
+  CARGO_TERM_COLOR: always
+  S3_BUCKET_NAME: ${{ vars.S3_EXPRESS_ONE_ZONE_BUCKET_NAME }}
+  S3_BUCKET_TEST_PREFIX: ${{ vars.S3_BUCKET_BENCH_PREFIX || 'mountpoint-benchmark/' }}
+  S3_BUCKET_BENCH_FILE: ${{ vars.BENCH_FILE_NAME || 'bench100GB.bin' }}
+  S3_BUCKET_SMALL_BENCH_FILE: ${{ vars.SMALL_BENCH_FILE_NAME || 'bench5MB.bin' }}
+  S3_REGION: ${{ vars.S3_REGION }}
+
+jobs:
+  bench:
+    name: Benchmark (Throughput)
+    runs-on: [self-hosted, linux, x64, high-performance]
+
+    environment: ${{ inputs.environment }}
+
+    steps:
+    - name: Configure AWS credentials
+      uses: aws-actions/configure-aws-credentials@v3
+      with:
+        role-to-assume: ${{ vars.ACTIONS_IAM_ROLE }}
+        aws-region: ${{ vars.S3_REGION }}
+        role-duration-seconds: 21600
+    - name: Checkout code
+      uses: actions/checkout@v3
+      with:
+        ref: ${{ inputs.ref }}
+        submodules: true
+        persist-credentials: false
+    - name: Install operating system dependencies
+      uses: ./.github/actions/install-dependencies
+      with:
+        fuseVersion: 2
+        libunwind: true
+        fio: true
+    - name: Set up stable Rust
+      uses: actions-rs/toolchain@v1
+      with:
+        toolchain: stable
+        override: true
+    - name: Restore Cargo cache
+      id: restore-cargo-cache
+      uses: actions/cache/restore@v3
+      with:
+        path: |
+          ~/.cargo/bin/
+          ~/.cargo/registry/index/
+          ~/.cargo/registry/cache/
+          ~/.cargo/git/db/
+          target/
+        key: ${{ runner.os }}-${{ github.job }}-cargo-${{ hashFiles('**/Cargo.lock') }}
+    - name: Build
+      run: cargo build --release
+    - name: Run Benchmark
+      run: mountpoint-s3/scripts/fs_bench.sh
+    - name: Save Cargo cache
+      uses: actions/cache/save@v3
+      if: inputs.environment != 'PR benchmarks'
+      with:
+        path: |
+          ~/.cargo/bin/
+          ~/.cargo/registry/index/
+          ~/.cargo/registry/cache/
+          ~/.cargo/git/db/
+          target/
+        key: ${{ steps.restore-cargo-cache.outputs.cache-primary-key }}
+    - name: Check benchmark results
+      uses: benchmark-action/github-action-benchmark@v1
+      with:
+        tool: 'customBiggerIsBetter'
+        output-file-path: results/output.json
+        benchmark-data-dir-path: dev/s3-express/bench
+        alert-threshold: "200%"
+        fail-on-alert: true
+        # GitHub API token to make a commit comment
+        github-token: ${{ secrets.GITHUB_TOKEN }}
+        # Store the results and deploy GitHub pages automatically if the results are from main branch
+        auto-push: ${{ inputs.environment && 'false' || 'true' }}
+        comment-on-alert: true
+        max-items-in-chart: 20
+
+  latency-bench:
+    name: Benchmark (Latency)
+    runs-on: [self-hosted, linux, x64, high-performance]
+
+    environment: ${{ inputs.environment }}
+
+    steps:
+    - name: Configure AWS credentials
+      uses: aws-actions/configure-aws-credentials@v3
+      with:
+        role-to-assume: ${{ vars.ACTIONS_IAM_ROLE }}
+        aws-region: ${{ vars.S3_REGION }}
+    - name: Checkout code
+      uses: actions/checkout@v3
+      with:
+        ref: ${{ inputs.ref }}
+        submodules: true
+        persist-credentials: false
+    - name: Install operating system dependencies
+      uses: ./.github/actions/install-dependencies
+      with:
+        fuseVersion: 2
+        libunwind: true
+        fio: true
+    - name: Set up stable Rust
+      uses: actions-rs/toolchain@v1
+      with:
+        toolchain: stable
+        override: true
+    - name: Restore Cargo cache
+      id: restore-cargo-cache
+      uses: actions/cache/restore@v3
+      with:
+        path: |
+          ~/.cargo/bin/
+          ~/.cargo/registry/index/
+          ~/.cargo/registry/cache/
+          ~/.cargo/git/db/
+          target/
+        key: ${{ runner.os }}-${{ github.job }}-cargo-${{ hashFiles('**/Cargo.lock') }}
+    - name: Build
+      run: cargo build --release
+    - name: Run Benchmark
+      run: mountpoint-s3/scripts/fs_latency_bench.sh
+    - name: Save Cargo cache
+      uses: actions/cache/save@v3
+      if: inputs.environment != 'PR benchmarks'
+      with:
+        path: |
+          ~/.cargo/bin/
+          ~/.cargo/registry/index/
+          ~/.cargo/registry/cache/
+          ~/.cargo/git/db/
+          target/
+        key: ${{ steps.restore-cargo-cache.outputs.cache-primary-key }}
+    - name: Check benchmark results
+      uses: benchmark-action/github-action-benchmark@v1
+      with:
+        tool: 'customSmallerIsBetter'
+        output-file-path: results/output.json
+        benchmark-data-dir-path: dev/s3-express/latency_bench
+        alert-threshold: "200%"
+        fail-on-alert: true
+        # GitHub API token to make a commit comment
+        github-token: ${{ secrets.GITHUB_TOKEN }}
+        # Store the results and deploy GitHub pages automatically if the results are from main branch
+        auto-push: ${{ inputs.environment && 'false' || 'true' }}
+        comment-on-alert: true
+        max-items-in-chart: 20

mountpoint-s3/scripts/fs_bench.sh

@@ -99,7 +99,8 @@ read_benchmark () {
       --debug \
       --allow-delete \
       --log-directory=${log_dir} \
-      --prefix=${S3_BUCKET_TEST_PREFIX}
+      --prefix=${S3_BUCKET_TEST_PREFIX} \
+      --part-size=16777216
     mount_status=$?
     set -e
     if [ $mount_status -ne 0 ]; then

mountpoint-s3/scripts/fs_latency_bench.sh

@@ -36,10 +36,13 @@ mkdir -p ${results_dir}
 
 # start readdir benchmarking
 dir_size=100
+jobs_dir=mountpoint-s3/scripts/fio/create
+
 while [ $dir_size -le 100000 ]
 do
     sum=0
     job_name="readdir_${dir_size}"
+    job_file="${jobs_dir}/create_files_${dir_size}.fio"
     mount_dir=$(mktemp -d /tmp/mount-s3-XXXXXXXXXXXX)
     target_dir="${mount_dir}/bench_dir_${dir_size}"
     startdelay=30
@@ -61,11 +64,9 @@ do
         exit 1
     fi
 
-    # verify that the target directory exists before running the benchmark
-    if [ ! -d "${target_dir}" ]; then
-      echo "Target directory ${target_dir} does not exist."
-      exit 1
-    fi
+    # create the target directory if it does not exist before running the benchmark
+    mkdir -p ${target_dir}
+    fio --directory=${target_dir} ${job_file}
 
     sleep $startdelay
     # run each case for 10 iterations