zb-system-tests-cloudbuild.yaml

substitutions:
  _REGION: "us-central1"
  _ZONE: "us-central1-a"
  _SHORT_BUILD_ID: ${BUILD_ID:0:8}
  _VM_NAME: "py-sdk-sys-test-${_SHORT_BUILD_ID}"
  _ULIMIT: "10000" # 10k, for gRPC bidi streams



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: ["-"]

  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
    id: "cleanup-old-keys"
    entrypoint: "bash"
    args:
      - "-c"
      - |
        #!/bin/bash
        set -e

        echo "Fetching OS Login SSH keys..."
        echo "Removing all keys."
        echo "---------------------------------------------------------------------"

        FINGERPRINTS_TO_DELETE=$$(gcloud compute os-login ssh-keys list \
          --format="value(fingerprint)")

        echo "Keys to delete: $$FINGERPRINTS_TO_DELETE"

        if [ -z "$$FINGERPRINTS_TO_DELETE" ]; then
          echo "No keys found to delete. Nothing to do."
          exit 0
        fi

        while IFS= read -r FINGERPRINT; do
          if [ -n "$$FINGERPRINT" ]; then
            echo "Deleting key with fingerprint: $$FINGERPRINT"
            gcloud compute os-login ssh-keys remove \
              --key="$$FINGERPRINT" \
              --quiet || true
          fi
        done <<< "$$FINGERPRINTS_TO_DELETE"

        echo "---------------------------------------------------------------------"
        echo "Cleanup complete."

  # 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=${_ZONE}"
      - "--machine-type=e2-medium"
      - "--image-family=debian-13"
      - "--image-project=debian-cloud"
      - "--service-account=${_ZONAL_VM_SERVICE_ACCOUNT}"
      - "--scopes=https://www.googleapis.com/auth/devstorage.full_control,https://www.googleapis.com/auth/devstorage.read_only,https://www.googleapis.com/auth/devstorage.read_write"
      - "--metadata=enable-oslogin=TRUE"
    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 main/storage/cloudbuild/run_zonal_tests.sh ${_VM_NAME}:~ --zone=${_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=${_ZONAL_BUCKET} CROSS_REGION_BUCKET=${_CROSS_REGION_BUCKET} _PR_NUMBER=${_PR_NUMBER} bash run_zonal_tests.sh"
        EXIT_CODE=$?
        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"
      - "cleanup-old-keys"

  - 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"