Gcp Serverless Fullstack Architecture
Use When
- GCP-first architecture with managed scale
- Event-driven workflows with Pub/Sub
Workflow
- Choose Cloud Run vs Cloud Functions per workload shape
- Define API Gateway or load balancer ingress and auth
- Implement services with clear contract versioning
- Configure Cloud CDN and frontend hosting strategy
- Use Pub/Sub with dead-letter topics and replay procedures
- Set SLOs, dashboards, and error budgets in Cloud Monitoring
Required Checks
- Service accounts follow least privilege per workload
- Ingress auth validated on every protected route
- Idempotency enforced for async consumers
- Multi-region or failover strategy documented where required
Decision Framework
- Prefer Cloud Run for containerized services with flexible runtime needs
- Use Pub/Sub for decoupled domain events
- Keep synchronous call graphs shallow for latency control
- Treat IaC as source of truth across environments
Common Rationalizations And Rebuttals
- "Cloud Run removes all backend concerns." -> Concurrency, cold start, and egress still need design.
- "Pub/Sub ordering is guaranteed globally." -> Design for duplicates and disorder.
- "Firebase rules replace API auth." -> Backend must enforce authorization independently.
Evidence Pack
- Auth negative test suite results
- Pub/Sub replay and poison-message handling proof
- CDN cache and purge validation during release
- Terraform plan and rollback checklist
Exit Criteria
- GCP serverless delivery is observable and recoverable
- Release can be rolled back without data corruption
