Production Readiness

Use this checklist before shipping Nimbus apps to production. It focuses on reliability, security, observability, and safe deployment workflows.

1) Core runtime safety

• Set APP_ENV=production and verify config is loaded from environment.
• Run behind a reverse proxy (TLS termination, forwarded headers, request limits).
• Ensure graceful shutdown is configured for rolling deploys.

2) Database & migrations

• Run migrations in CI/staging before production rollout.
• Keep migrations small and reversible (Up/Down).
• For DDL that cannot run in transactions, mark migration as non-transactional and document rollback path.
• Never edit an already-applied migration in production.

Related: Migrations, Database

3) Queue reliability

• Use QUEUE_DRIVER=redis or database (not sync) in production.
• Set QUEUE_BOOT_STRICT=true so invalid queue config fails at boot.
• Set QUEUE_REDIS_VISIBILITY_TIMEOUT_SECONDS / QUEUE_DB_LEASE_SECONDS to at least 2x p95 job runtime.
• Set bounded retries and keep job handlers idempotent.
• Run dedicated worker processes and monitor backlog growth.

Related: Queue, Horizon

4) Realtime security

• Configure explicit WebSocket / Presence / Reverb origin allowlists (REVERB_ALLOWED_ORIGINS).
• Authorize channels in Presence via AuthFunc.
• Avoid broad CORS rules unless absolutely necessary.

Related: WebSockets, Reverb, Presence, CORS

5) Health checks & observability

• Expose health endpoints for load balancers and orchestration probes (/livez, /readyz, legacy /health).
• Track queue counters (dispatch/processed/failed/retried/reclaimed).
• Alert on retry spikes, reclaim spikes, and sustained queue backlog.

# Horizon — Prometheus + JSON workloads (when RedisURL configured)
GET /horizon/api/metrics/prometheus
GET /horizon/api/workloads

# Reverb — liveness next to WebSocket path
GET /reverb/health

Related: Health, Metrics

6) Release process

• Run CI gates on every PR: tests, race detector, and go vet.
• Require peer review for migration and queue-related changes.
• Roll out with staged deploys and rollback plan validated.

Pre-launch quick check

[ ] APP_ENV=production configured
[ ] Migrations tested in staging and rollback verified
[ ] Queue driver + lease timeouts configured
[ ] WebSocket/Presence allowed origins configured
[ ] Health and metrics endpoints reachable
[ ] Alerts configured for queue failures/retries/reclaims
[ ] CI checks passing on default branch
[ ] Deployment rollback plan documented