Free resource
The Laravel Scaling Checklist
Twenty six checks I run before calling a Laravel application production ready. Use it as a pre-launch review, or as an audit of a system that is already struggling under load.
Want it as a PDF, plus new checklists?
Drop your email and I will send the printable version. No newsletter spam, only practical backend material.
Your email is used for this request only and never shared.
1. Database and queries
- Every query used on a hot path has a covering index, verified with EXPLAIN.
- N+1 queries are eliminated with eager loading and caught in CI by a query counter.
- Slow query log is enabled with a 200ms threshold and reviewed weekly.
- Read-heavy reporting runs against a replica, never the primary.
- Large tables have a retention or partitioning plan before they pass 50 million rows.
2. Caching
- Redis is used for sessions, rate limits, counters, and precomputed dashboard tiles.
- Cache keys carry a version prefix so a deploy can invalidate a whole class of keys.
- TTLs have jitter and expensive keys are rebuilt behind Cache::lock to avoid stampedes.
- Config, routes, views, and events are cached in production builds.
3. Queues and background work
- Nothing that can fail or take longer than 200ms happens inside a web request.
- Horizon is running with separate queues for critical, default, and bulk work.
- Every job is idempotent and safe to retry, with backoff configured.
- Failed jobs alert the channel the team already watches, not an unread inbox.
4. API and integrations
- All external calls have timeouts, retries with backoff, and a circuit breaker.
- Webhooks verify signatures and are processed asynchronously.
- Rate limiting is applied per client, not globally.
- API responses are versioned and paginated by cursor on large collections.
5. Observability
- Structured logs with a request ID that follows a job into the queue.
- Error tracking with release tagging so regressions are attributed to a deploy.
- Dashboards for queue depth, job runtime, database connections, and p95 latency.
- Alerts are based on user-visible symptoms, not CPU graphs.
6. Security and reliability
- Authorization enforced by policies, and tenancy enforced at the query layer.
- Secrets live in the environment, never in the repository.
- Automated daily backups with a restore that has actually been tested.
- Zero-downtime deploys with migrations that are backwards compatible.
Failing more than five of these?
Book a free 30 minute review. We go through your architecture together and you leave with the shortest path to a fix.
Book a free review