A replica will not start: "REDIS_URL is required" or "Redis is invalid or unreachable"

In production, Hardcount uses Redis for rate limiting and refuses to start without it. Back to the runbook index.

Symptom

A production replica refuses to boot and logs one of these messages:

• REDIS_URL is required in production
• refusing to start: REDIS_URL is set but Redis is invalid or unreachable

Why

The general and hosted-invoice rate limiters fail closed without Redis. A replica without Redis would boot green and answer 429 to its share of traffic, including hosted-invoice pay pages, while /health/ready stays ok. Refusing to start is the safer outcome.

Check

• Is REDIS_URL set, and does it point at the right host?
• Is the network path to port 6379 open (security group, firewall)?
• Is the managed Redis up?

Fix

Correct REDIS_URL, the network path, or the Redis service, then start the replica again.

Verify

The replica starts on the next attempt.

Redis going down after boot

This is a different situation:

• General and hosted-invoice requests answer 429 (they fail closed).
• Usage ingest keeps working with no limit (it fails open).
• Sign-in, /v1/auth, keeps working and stays limited. Each replica counts sign-in requests on its own. The limit is 100 a minute per address by default, so with three replicas one address can make up to 300 a minute. The log shows rate_limiter: redis error, counting in this process.
• Restore Redis. No restart is needed.