CI/CD Pipeline¶

The CI/CD pipeline is defined in .github/workflows/deploy.yml and .github/workflows/seed-mind-data.yml.

Performance Optimizations (FAS-79)¶

The pipeline includes several optimizations for faster builds:

1. Next.js build caching: Test job caches .next for Docker build (saves ~2-3 min)
2. Skip duplicate build: Test job skips build on manual deploy (saves ~3 min)
3. Reduced BuildKit cache: Changed from mode=max to mode=min (saves ~5-6 min)
4. Skip lint/type-check in Docker: Uses build args since test job already validated (saves ~45-50s)
5. Improved Dockerfile layering: Better cache invalidation with npm cache mounts (saves ~30-60s)
6. Separate MIND seeding workflow: Long-running data import moved to separate workflow (saves ~10+ min on deploys)

Total expected savings: 10-15 minutes on average builds

Triggers¶

The GitHub OIDC role trusts all branches (refs/heads/*), enabling team members to run CI/CD from feature branches using workflow_dispatch.

Important: Push to main no longer triggers automatic deployment. Use manual workflow dispatch to deploy.

Pipeline Jobs¶

detect-changes¶

Determines what needs to run based on file changes.

• Detects app changes (src, prisma, Dockerfile, etc.)
• Detects infra changes (infra/, .github/workflows/)
• Decides whether to build and/or deploy
• Skips unnecessary steps for performance

test¶

Runs on every PR, push to main, and manual deploy.

steps:
  - npm ci
  - npm run lint
  - npm run type-check
  - npm run test
  - npm run build (skipped on manual deploy - Docker will build)
  - Cache .next for Docker build (PR/push only)

Optimization: On manual deploy, build is skipped to avoid duplication. Docker build reuses the cached .next from PR/push.

terraform-plan (PR only)¶

Shows infrastructure changes without applying them.

• Runs terraform plan
• Posts plan output as PR comment
• Helps review infrastructure changes

ensure-ecr (manual deploy only)¶

Ensures ECR repository exists before pushing images.

build-and-push (manual deploy only)¶

Builds and pushes Docker image to ECR.

• Platform: linux/amd64
• Tag: sha-{commit}-run-{run_number}
• Uses BuildKit cache (mode=min for faster builds)
• Restores Next.js build cache from test job
• Passes SKIP_LINT=true and SKIP_TYPE_CHECK=true to Docker (already validated in test job)

deploy (manual deploy only)¶

Deploys to App Runner.

steps:
  - terraform init
  - terraform apply
  - Wait for deployment (up to 5 minutes)
  - Health check (up to 15 retries)

Required Secrets¶

Deployment Process¶

1. Quality gates pass - lint, type-check, test
2. ECR repository ready - Created if needed
3. Docker image built - Multi-stage build with caching, pushed to ECR
4. Terraform apply - Infrastructure updated (reuses existing image if infra-only change)
5. Database migrations - Run against RDS with automatic failed migration resolution
6. Wait for deployment - App Runner updates service
7. Health check - Verifies /api/health returns 200

MIND UAT Data Seeding¶

MIND test data seeding has been separated into its own workflow (.github/workflows/seed-mind-data.yml) for performance reasons. To seed test data:

1. Go to Actions → "Seed MIND UAT Data"
2. Click "Run workflow"
3. Select environment (currently only dev)
4. Choose whether to force re-seed (overwrites existing data)
5. Click "Run workflow"

This workflow: - Connects to RDS database - Runs npm run db:seed:mind - Imports data from three commodity verticals (Cobalt, LBR SYP Lumber, South America Vegoils) - Takes ~10+ minutes to complete

Viewing Pipeline Status¶

GitHub Actions UI¶

1. Go to repository → Actions tab
2. Click on workflow run
3. View job logs

PR Comments¶

Terraform plan output is posted as a PR comment:

## Terraform Plan ✅ Plan Succeeded

<details>
<summary>Click to expand</summary>

... plan output ...

</details>

Troubleshooting¶

Build Failures¶

Lint errors:

npm run lint
# Fix reported issues

Type errors:

npm run type-check
# Fix reported issues

Build errors:

npm run build
# Check for missing dependencies or configuration issues

Deployment Failures¶

Terraform errors: - Check Terraform plan in PR comments - Verify secrets are set correctly - Check AWS permissions

Health check failures: - Check App Runner logs in AWS Console - Verify DATABASE_URL is correct - Check application logs

Manual Deployment¶

You can trigger a manual deployment from any branch:

1. Go to Actions → "Luminarium Proof CI/CD"
2. Click "Run workflow"
3. Select your branch (e.g., main or a feature branch)
4. Choose deployment options:
5. environment: Target environment - dev or test (default: dev)
6. deploy: Deploy to target environment (default: true)
7. force_build: Rebuild image even without app changes (default: false)
8. Click "Run workflow"

This is useful for: - Testing infrastructure changes from feature branches - Forcing a rebuild when needed - Deploying hotfixes