fcsouza/infrastructure-ci-cd-game

CI/CD for Games

Purpose

GitHub Actions pipelines, test strategies, migration safety, and staging environments for game projects.

When to Use

Trigger: CI/CD, GitHub Actions, deployment, pipeline, testing, staging, migration, continuous integration, continuous deployment, build pipeline, release

Prerequisites

• postgres-game-schema — migration safety
• game-backend-architecture — what we're deploying

Core Principles

1. Never deploy without tests — automated tests gate every deployment
2. Migrations are one-way — never run destructive migrations in production without backup
3. Staging mirrors production — test with real data shapes, real Redis, real PostgreSQL
4. Feature flags over branches — deploy dark features, enable via config
5. Rollback plan always — every deployment has a revert strategy
6. Monitor after deploy — monitoring-game-ops alerts catch post-deploy issues

Pipeline Stages

Push → Lint → Type Check → Unit Tests → Build → Integration Tests → Deploy Staging → Smoke Tests → Deploy Production

Step-by-Step: CI Pipeline

Set up CI with Bun + GitHub Actions:

1. Use oven-sh/setup-bun@v2 — not actions/setup-node. Bun is the runtime.
2. Cache bun install:

   - uses: actions/cache@v4
     with:
       path: ~/.bun/install/cache
       key: ${{ runner.os }}-bun-${{ hashFiles('**/bun.lockb') }}
       restore-keys: |
         ${{ runner.os }}-bun-
   

3. Steps in order:
   • bun install --frozen-lockfile — reproducible installs
   • bunx biome check . — lint and format check
   • bun run typecheck — TypeScript strict mode
   • bun test — unit and integration tests
4. Environment variables for tests:
   • DATABASE_URL — Neon test branch connection string
   • REDIS_URL — test Redis instance

See boilerplate/ci.yml for the complete workflow.

Step-by-Step: Deploy Pipeline

1. Drizzle migration job (runs before deploy):
   • Separate job that runs bunx drizzle-kit migrate with production DATABASE_URL
   • Deploy job declares needs: migrate so it only runs after migrations succeed
2. Deploy to Fly.io:
   • Use superfly/flyctl-actions/setup-flyctl@master to install flyctl
   • Run flyctl deploy --remote-only — builds on Fly's infrastructure, no local Docker required
   • Auth via FLY_API_TOKEN secret
3. Post-deploy health check:

   APP_NAME=$(flyctl info --json | jq -r '.Name')
   curl --retry 5 --retry-delay 10 -f "https://${APP_NAME}.fly.dev/health"
   

4. Rollback strategy:
   • Fly.io machines support instant rollback: flyctl releases rollback
   • Every deploy creates a new release — previous release is always available
   • Combine with feature flags to disable broken features without a full rollback

See boilerplate/deploy.yml for the complete workflow.

Neon Database Branching

For preview environments (per-PR Neon branching):

• Neon supports database branches — each PR gets its own isolated database branch
• Use neondatabase/create-branch-action@v5 in GitHub Actions:

  - uses: neondatabase/create-branch-action@v5
    id: create_branch
    with:
      project_id: ${{ secrets.NEON_PROJECT_ID }}
      branch_name: preview/pr-${{ github.event.number }}
      api_key: ${{ secrets.NEON_API_KEY }}
  

• Branch connection string available as ${{ steps.create_branch.outputs.db_url }}
• Delete branch on PR close with neondatabase/delete-branch-action@v3
• This gives each PR a real isolated database — no shared test state pollution

boilerplate/ci.yml and boilerplate/deploy.yml

See boilerplate/ci.yml for a complete, production-ready CI workflow and boilerplate/deploy.yml for the deployment workflow with migration safety. Both files are ready to copy into .github/workflows/.

Cross-References

• postgres-game-schema — migration safety in deployments
• monitoring-game-ops — post-deployment monitoring
• game-backend-architecture — server deployment targets

Sources

• GitHub Actions documentation
• "Continuous Delivery for Games" — GDC DevOps talks