feat: run benchmarks in CI

cloudbuild/backup.yaml

@@ -0,0 +1,19 @@
+gcloud compute instances create 
+py-sdk-micro-benchmarks-ci-test-2 
+--project=gcs-aiml-clients-testing-101 
+--zone=us-central1-a 
+--machine-type=c4-standard-192 
+--network-interface=network-tier=PREMIUM,nic-type=GVNIC,stack-type=IPV4_ONLY,subnet=default 
+--metadata=enable-osconfig=TRUE,enable-oslogin=TRUE
+--maintenance-policy=MIGRATE
+--provisioning-model=STANDARD
+--service-account=gcsfs-perf-ci-runner@gcs-aiml-clients-testing-101.iam.gserviceaccount.com
+--scopes=https://www.googleapis.com/auth/cloud-platform,https://www.googleapis.com/auth/devstorage.read_write
+--tags=allow-ssh,http-server,https-server,lb-health-check
+--create-disk=auto-delete=yes,boot=yes,device-name=py-sdk-micro-benchmarks-ci-test,disk-resource-policy=projects/gcs-aiml-clients-testing-101/regions/us-central1/resourcePolicies/default-schedule-1,image=projects/ubuntu-os-cloud/global/images/ubuntu-2204-jammy-v20251120,mode=rw,provisioned-iops=3600,provisioned-throughput=290,size=100,type=hyperdisk-balanced
+--no-shielded-secure-boot
+--shielded-vtpm
+--shielded-integrity-monitoring
+--labels=goog-ops-agent-policy=v2-x86-template-1-4-0,goog-ec-src=vm_add-gcloud
+--reservation-affinity=any
+--network-performance-configs=total-egress-bandwidth-tier=TIER_1

cloudbuild/micro_benchmarks.yaml

@@ -0,0 +1,111 @@
+substitutions:
+  _REGION: "us-central1"
+  _ZONE: "us-central1-a"
+  _SHORT_BUILD_ID: ${BUILD_ID:0:8}
+  _VM_NAME: "py-sdk-micro-benchmarks-ci-${_SHORT_BUILD_ID}"
+  _ULIMIT: "10000" # 10k, for gRPC bidi streams
+  _BENCHMARK_SCRIPT: "run_micro_benchmarks.sh"
+
+
+
+steps:
+  # Step 0: Generate a persistent SSH key for this build run.
+  # This prevents gcloud from adding a new key to the OS Login profile on every ssh/scp command.
+  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
+    id: "generate-ssh-key"
+    entrypoint: "bash"
+    args:
+      - "-c"
+      - |
+        mkdir -p /workspace/.ssh
+        # Generate the SSH key
+        ssh-keygen -t rsa -f /workspace/.ssh/google_compute_engine -N '' -C gcb
+        # Save the public key content to a file for the cleanup step
+        cat /workspace/.ssh/google_compute_engine.pub > /workspace/gcb_ssh_key.pub
+    waitFor: ["-"]
+
+  # Step 1 Create a GCE VM to run the tests.
+  # The VM is created in the same zone as the buckets to test rapid storage features.
+  # It's given the 'cloud-platform' scope to allow it to access GCS and other services.
+  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
+    id: "create-vm"
+    entrypoint: "gcloud"
+    args:
+      - "compute"
+      - "instances"
+      - "create"
+      - "${_VM_NAME}"
+      - "--project=${PROJECT_ID}"
+      - "--zone=${_ZONE}"
+      - "--machine-type=c4-standard-192"
+      - "--network-interface=network-tier=PREMIUM,nic-type=GVNIC,stack-type=IPV4_ONLY,subnet=default"
+      - "--metadata=enable-osconfig=TRUE,enable-oslogin=TRUE"
+      - "--maintenance-policy=MIGRATE"
+      - "--provisioning-model=STANDARD"
+      - "--service-account=${_ZONAL_VM_SERVICE_ACCOUNT}"
+      - "--scopes=https://www.googleapis.com/auth/cloud-platform,https://www.googleapis.com/auth/devstorage.read_write"
+      - "--tags=allow-ssh,http-server,https-server,lb-health-check"
+      - "--create-disk=auto-delete=yes,boot=yes,device-name=${_VM_NAME},disk-resource-policy=projects/${PROJECT_ID}/regions/${_REGION}/resourcePolicies/default-schedule-1,image=projects/ubuntu-os-cloud/global/images/ubuntu-2204-jammy-v20251120,mode=rw,provisioned-iops=3600,provisioned-throughput=290,size=100,type=hyperdisk-balanced"
+      - "--no-shielded-secure-boot"
+      - "--shielded-vtpm"
+      - "--shielded-integrity-monitoring"
+      - "--labels=goog-ops-agent-policy=v2-x86-template-1-4-0,goog-ec-src=vm_add-gcloud"
+      - "--reservation-affinity=any"
+      - "--network-performance-configs=total-egress-bandwidth-tier=TIER_1"
+    waitFor: ["-"]
+
+  # Step 2: Run the integration tests inside the newly created VM and cleanup.
+  # This step uses 'gcloud compute ssh' to execute a remote script.
+  # The VM is deleted after tests are run, regardless of success.
+  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
+    id: "run-tests-and-delete-vm"
+    entrypoint: "bash"
+    args:
+      - "-c"
+      - |
+        set -e
+        # Wait for the VM to be fully initialized and SSH to be ready.
+        for i in {1..10}; do
+          if gcloud compute ssh ${_VM_NAME} --zone=${_ZONE} --internal-ip --ssh-key-file=/workspace/.ssh/google_compute_engine --command="echo VM is ready"; then
+            break
+          fi
+          echo "Waiting for VM to become available... (attempt $i/10)"
+          sleep 15
+        done
+        # copy the script to the VM
+        gcloud compute scp cloudbuild/${_BENCHMARK_SCRIPT} ${_VM_NAME}:~ --zone=${_ZONE} --internal-ip --ssh-key-file=/workspace/.ssh/google_compute_engine
+
+        # Execute the script on the VM via SSH.
+        # Capture the exit code to ensure cleanup happens before the build fails.
+        set +e
+        gcloud compute ssh ${_VM_NAME} --zone=${_ZONE} --internal-ip --ssh-key-file=/workspace/.ssh/google_compute_engine --command="ulimit -n {_ULIMIT}; COMMIT_SHA=${COMMIT_SHA} _ZONAL_BUCKET=${_ZONAL_BUCKET} _PR_NUMBER=${_PR_NUMBER} bash ${_BENCHMARK_SCRIPT}"
+        EXIT_CODE=$?
+        set -e
+
+        echo "--- Deleting GCE VM ---"
+        gcloud compute instances delete "${_VM_NAME}" --zone=${_ZONE} --quiet
+
+        # Exit with the original exit code from the test script.
+        exit $$EXIT_CODE
+    waitFor:
+      - "create-vm"
+      - "generate-ssh-key"
+
+  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
+    id: "cleanup-ssh-key"
+    entrypoint: "bash"
+    args:
+      - "-c"
+      - |
+        echo "--- Removing SSH key from OS Login profile to prevent accumulation ---"
+        gcloud compute os-login ssh-keys remove \
+          --key-file=/workspace/gcb_ssh_key.pub || true
+    waitFor:
+      - "run-tests-and-delete-vm"
+
+timeout: "3600s" # 60 minutes
+
+options:
+  logging: CLOUD_LOGGING_ONLY
+  pool:
+    name: "projects/${PROJECT_ID}/locations/us-central1/workerPools/cloud-build-worker-pool"

cloudbuild/run_micro_benchmarks.sh

@@ -0,0 +1,32 @@
+
+set -euxo pipefail
+echo '--- Installing git and cloning repository on VM ---'
+sudo apt-get update && sudo apt-get install -y git python3-pip python3-venv
+
+# Clone the repository and checkout the specific commit from the build trigger.
+git clone https://github.com/googleapis/python-storage.git
+cd python-storage
+git fetch origin "refs/pull/${_PR_NUMBER}/head"
+git checkout ${COMMIT_SHA}
+
+
+echo '--- Installing Python and dependencies on VM ---'
+python3 -m venv env
+source env/bin/activate
+
+echo 'Install testing libraries explicitly, as they are not in setup.py'
+pip install --upgrade pip
+pip install pytest pytest-timeout pytest-subtests pytest-asyncio
+pip install google-cloud-testutils google-cloud-kms
+pip install -e ".[grpc,testing]"
+
+echo '--- Setting up environment variables on VM ---'
+export ZONAL_BUCKET=${_ZONAL_BUCKET}
+export RUN_ZONAL_SYSTEM_TESTS=True
+CURRENT_ULIMIT=$(ulimit -n)
+echo '--- Running Zonal tests on VM with ulimit set to ---' $CURRENT_ULIMIT
+pytest -vv -s \
+    --log-format='%(asctime)s %(levelname)s %(message)s' \
+    --log-date-format='%H:%M:%S' \
+    --benchmark-json=output.json \
+    -vv -s tests/perf/microbenchmarks/reads/test_reads.py::test_downloads_single_proc_single_coro

tests/perf/microbenchmarks/reads/config.yaml

@@ -1,11 +1,11 @@
 common:
   bucket_types:
-    - "regional"
+    # - "regional"
     - "zonal"
   file_sizes_mib:
-    - 1024 # 1GiB
-  chunk_sizes_mib: [100]
-  rounds: 10
+    - 100 # 1GiB
+  chunk_sizes_mib: [10]
+  rounds: 3
 
 workload: