Production Launch Checklist
Every item to verify before going live — security, performance, monitoring, backups, and the final pre-launch walkthrough
Skills taught: deployment.production-checklist, deployment.security-headers, deployment.launch-readiness, deployment.performance-baseline
Why a checklist?
Aviation uses checklists not because pilots are incompetent, but because humans forget things under pressure. Launching a product involves dozens of steps, many of which interact with each other. A single missed step — HTTPS redirect not enabled, error tracking not configured, backups pointing at the wrong database — can cost you users, data, or credibility. This checklist covers everything we've learned in this track, organized as a pre-launch runthrough.
Security checklist
HTTPS:
[ ] HTTPS enabled — no HTTP-only responses
[ ] HTTP → HTTPS redirect (301) is configured
[ ] HSTS header set (Strict-Transport-Security: max-age=31536000)
[ ] Certificate is valid and auto-renews
HEADERS:
[ ] X-Frame-Options: DENY (or SAMEORIGIN if embedding needed)
[ ] X-Content-Type-Options: nosniff
[ ] Content-Security-Policy configured (prevents XSS injection)
[ ] Referrer-Policy: strict-origin-when-cross-origin
[ ] Permissions-Policy set (disable unused browser features)
SECRETS:
[ ] No secrets in git history (check: git log -p | grep -i secret)
[ ] .env NOT committed (check: git ls-files | grep .env)
[ ] .env.example committed with empty values
[ ] Production secrets stored in platform secrets manager
[ ] Different secrets for staging vs production
AUTHENTICATION:
[ ] Password hashed with bcrypt/argon2 (NEVER plain text or MD5)
[ ] JWT secrets are long (32+ random bytes), not 'mysecret'
[ ] Session tokens are httpOnly cookies (not localStorage for auth)
[ ] Rate limiting on login endpoint (prevent brute force)
[ ] CSRF protection on state-changing endpoints
DATABASE:
[ ] Database not exposed to internet (only accessible from app server)
[ ] DB user has minimum required permissions (not superuser)
[ ] SQL queries use parameterized queries (prevent SQL injection)
Performance checklist
STATIC ASSETS:
[ ] Gzip/Brotli compression enabled on Nginx or CDN
[ ] CSS and JS minified (Vite does this in production build)
[ ] Images compressed and in modern format (WebP, AVIF)
[ ] Cache-Control: immutable, max-age=31536000 on hashed assets
[ ] Cache-Control: no-cache on index.html
CDN:
[ ] Static files served from CDN (Cloudflare, CloudFront, Vercel Edge)
[ ] CDN configured for your domain (not origin-only serving)
DATABASE:
[ ] Indexes on all foreign keys
[ ] Indexes on columns used in WHERE clauses (check query plan)
[ ] N+1 query problem absent (use eager loading, not per-row queries)
[ ] Database connection pooling enabled
PAGE SPEED:
[ ] Lighthouse score: Performance > 80 (run: npm run build && serve -s dist)
[ ] LCP (Largest Contentful Paint) < 2.5 seconds
[ ] CLS (Cumulative Layout Shift) < 0.1
[ ] Bundle size checked: npm run build (Vite shows sizes)
[ ] No unused dependencies included in bundle
Monitoring and reliability checklist
UPTIME:
[ ] Uptime monitoring configured (BetterStack, UptimeRobot)
[ ] Health check endpoint (/health) responding with 200
[ ] Alert goes to email AND Slack (or phone for critical apps)
[ ] Status page configured (optional but professional)
ERROR TRACKING:
[ ] Sentry (or equivalent) initialized in both frontend and backend
[ ] Error alerts configured (notify on new issues)
[ ] Source maps uploaded for readable stack traces
[ ] User context set in Sentry (so you know which users are affected)
LOGGING:
[ ] Structured JSON logging enabled
[ ] Logs being stored somewhere queryable (not just stdout)
[ ] Request IDs added to all log lines
[ ] Sensitive data NOT being logged (passwords, tokens, card numbers)
[ ] Log retention policy set (auto-delete after 30-90 days)
BACKUPS:
[ ] Automated database backup scheduled (at minimum: daily)
[ ] Backup restoration has been TESTED (not just configured)
[ ] Backup files stored outside the production server
[ ] Recovery runbook documented
Security headers implementation
import helmet from 'helmet'; // npm install helmet
// Helmet sets most security headers automatically
app.use(helmet({
contentSecurityPolicy: {
directives: {
defaultSrc: ["'self'"],
scriptSrc: ["'self'", "'unsafe-inline'"], // adjust for your CDN
styleSrc: ["'self'", "'unsafe-inline'", "https://fonts.googleapis.com"],
fontSrc: ["'self'", "https://fonts.gstatic.com"],
imgSrc: ["'self'", "data:", "https:"],
connectSrc: ["'self'", "https://api.myapp.com"],
},
},
// Prevents clickjacking
frameguard: { action: 'deny' },
// Forces HTTPS for 1 year
hsts: { maxAge: 31536000, includeSubDomains: true },
}));
// Verify headers at: https://securityheaders.com
// Also: https://observatory.mozilla.org/analyze/myapp.com
Day-of-launch checklist
BEFORE LAUNCH (1-2 days before):
[ ] Deploy to staging and test all critical flows manually
[ ] Run Lighthouse audit on staging
[ ] Confirm all environment variables are set in production
[ ] Test sending an alert (verify Slack/email notifications work)
[ ] Take a database backup and test restoration
[ ] Review error tracking — resolve any pre-existing errors
LAUNCH DAY:
[ ] Deploy latest build to production
[ ] Verify health check is green
[ ] Walk through critical user flows:
- Sign up / login
- Core feature (make a purchase, submit a form, etc.)
- Error state (invalid input, 404 page)
[ ] Check error tracking: no new errors immediately post-deploy
[ ] Check uptime monitoring: green status
[ ] Verify DNS, HTTPS, redirects all working
POST-LAUNCH (24 hours after):
[ ] Review error tracking for any new issues users encountered
[ ] Review logs for any suspicious patterns
[ ] Check database query performance (any slow queries?)
[ ] Confirm backups have been running
[ ] Take note of real traffic patterns for capacity planning
EMERGENCY ROLLBACK:
[ ] Documented: how to roll back to previous version
[ ] Previous version's Docker image or git tag recorded
[ ] Everyone who can deploy knows how to roll back
Tip: Check your security headers at securityheaders.com after launch. Enter your domain and get an instant grade (A+ to F) with explanations for what's missing. Mozilla Observatory (observatory.mozilla.org) gives a more detailed security scan including CORS policy, cookie settings, and CSP analysis. Both are free. An A or A+ grade means your headers are correctly configured. Aim for at least a B before launch.
Frequently Asked Questions
How long should a production launch take with this checklist?
For a new app: the security and performance items take 2-4 hours to implement the first time. After that, subsequent launches take 30 minutes for the day-of checklist. The monitoring and backup setup is one-time work. For an experienced team, everything on this checklist is standard practice that's built into the development workflow — you're not adding it at launch, you're verifying it's in place.