Deployment Runbook

Step-by-step procedures for deploying, verifying, and rolling back the booking platform.

Pre-Deployment Checklist

• [ ] All CI checks pass on the target branch (bun run verify)
• [ ] Database migrations reviewed and tested on staging
• [ ] Environment variables confirmed in Cloudflare Workers dashboard
• [ ] CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID set in GitHub Secrets
• [ ] Sentry release created and source maps uploaded
• [ ] Stakeholders notified of deployment window
• [ ] Rollback plan reviewed

Deployment Steps

1. Run Database Migrations

bun run --cwd packages/database drizzle-kit migrate

Verify migration output for errors before proceeding.

2. Deploy Website

bun run --cwd apps/website deploy

3. Deploy Admin

bun run --cwd apps/admin deploy

4. Deploy Docs (if changed)

bun run build:docs
bun run --cwd docs deploy

5. Verify Deployment

Check each endpoint:

curl -sS https://booking.example.com/.health
curl -sS https://admin.booking.example.com/.health
curl -sS https://docs.booking.example.com/.health

All endpoints must return 200 OK with {"status":"healthy"}.

6. Smoke Test Critical Flows

• Customer signup and login
• Provider search and booking creation
• Stripe checkout (test mode)
• Admin dashboard loads
• Chat message send and receive

Post-Deployment Verification

• [ ] Health endpoints return healthy
• [ ] Sentry shows no new errors
• [ ] Cloudflare Analytics shows normal traffic
• [ ] No alerts triggered (check Slack #alerts channel if configured)
• [ ] Smoke tests pass
• [ ] Deployment announcement sent to stakeholders

Rollback Procedure

Instant Rollback via Wrangler

wrangler deployments rollback --name booking-website
wrangler deployments rollback --name booking-admin

Database Rollback

If a migration caused issues:

1. Identify the problematic migration
2. Create a new down migration in packages/database/src/adapters/migrations/
3. Run bun run --cwd packages/database drizzle-kit migrate
4. Verify data integrity

Rollback Verification

• [ ] Health endpoints return healthy
• [ ] Sentry error rate returns to baseline
• [ ] Critical flows functional
• [ ] Stakeholders notified of rollback

Emergency Contacts

Role	Contact	Escalation
On-call engineer	See PagerDuty schedule	If unreachable within 15 min, escalate to secondary
Engineering lead	See team directory	For decisions affecting data integrity
Cloudflare support	Cloudflare dashboard	For platform-level incidents

Deployment Window

• Standard: Tuesday–Thursday 10:00–16:00 UTC
• Emergency: Anytime with on-call approval
• Freeze: During major holidays and announced freeze periods