Google Cloud Run Deployment

Deploy NodeTool workflows to Google Cloud Run for scalable, serverless execution.

Deploy NodeTool to Google Cloud Run for scalable, serverless workflow execution. Cloud Run automatically scales instances based on demand and you only pay for actual compute time.

Prerequisites

Before deploying, ensure you have:

  • Google Cloud SDK (gcloud) installed and authenticated
  • Docker installed for building container images
  • A Google Cloud project with billing enabled
  • The following APIs enabled in your project:
    • Cloud Run API
    • Artifact Registry API (or Container Registry)
    • Cloud Build API (if using remote builds)

Enable Required APIs

gcloud services enable run.googleapis.com
gcloud services enable artifactregistry.googleapis.com
gcloud services enable cloudbuild.googleapis.com

Authenticate

gcloud auth login
gcloud auth configure-docker

Set your project:

gcloud config set project YOUR_PROJECT_ID

Or use the environment variable:

export GOOGLE_CLOUD_PROJECT=your-project-id

Quick Start

# 1. Authenticate with Google Cloud
gcloud auth login

# 2. Create a GCP deployment target
nodetool deploy add my-gcp --type gcp

# 3. Edit deployment.yaml to configure your settings (see below)

# 4. Deploy
nodetool deploy apply my-gcp

Configuration

The GCP deployment target supports these configuration fields in deployment.yaml:

Service Settings

Field Description Default
service_name Cloud Run service name (max 63 chars, alphanumeric + hyphens) nodetool
region GCP region (e.g., us-central1, europe-west1) us-central1
registry Artifact Registry URL (auto-inferred from region if omitted) –

Compute Resources

Field Description Default
cpu CPU allocation (e.g., 2, 4, 8) 2
memory Memory allocation (e.g., 4Gi, 8Gi, 16Gi) 4Gi
gpu_type GPU type (e.g., nvidia-l4) –
gpu_count Number of GPUs 0

Scaling

Field Description Default
min_instances Minimum running instances (0 = scale to zero) 0
max_instances Maximum instances for autoscaling 10
concurrency Max concurrent requests per instance 80
timeout Request timeout in seconds 300

Security

Field Description Default
service_account GCP service account email –
allow_unauthenticated Allow public access without auth false

Storage

Field Description Default
gcs_bucket Google Cloud Storage bucket for assets –
gcs_mount_path Mount path for GCS bucket in the container –

Environment Variables

Add provider API keys and configuration via the env section:

type: gcp
service_name: nodetool-prod
region: us-central1
cpu: 4
memory: 8Gi
min_instances: 1
max_instances: 5
env:
  AUTH_PROVIDER: supabase
  SUPABASE_URL: https://your-project.supabase.co
  SUPABASE_KEY: your-service-role-key
  OPENAI_API_KEY: sk-...

Operational Commands

# View deployment status
nodetool deploy status my-gcp

# Stream logs from running service
nodetool deploy logs my-gcp

# Preview changes without deploying
nodetool deploy plan my-gcp

# View current configuration
nodetool deploy show my-gcp

# Tear down the deployment
nodetool deploy destroy my-gcp

GPU Workflows on Cloud Run

Cloud Run supports GPU workloads for AI-intensive workflows:

  1. Request GPU access for your GCP project (may require quota increase)
  2. Configure GPU in your deployment: 
    gpu_type: nvidia-l4
gpu_count: 1
memory: 16Gi
  3. Deploy: nodetool deploy apply my-gcp

GPU instances have higher costs and may have limited region availability. Check Cloud Run GPU pricing for current rates.

Troubleshooting

Common Issues

“Permission denied” errors

  • Ensure your account has the Cloud Run Admin and Artifact Registry Writer roles
  • Check that all required APIs are enabled

Image push failures

  • Run gcloud auth configure-docker to set up Docker authentication
  • Verify Artifact Registry is enabled in your project

Service won’t start

  • Check logs: nodetool deploy logs my-gcp
  • Verify environment variables are set correctly
  • Ensure the container image builds successfully locally first

Cold start latency

  • Set min_instances: 1 to keep at least one instance warm
  • Use smaller container images when possible