# What Production-Ready Actually Means

> Six questions specific enough that two people would reach the same verdict.

Source: https://christwentyfour.com/blog/what-production-ready-actually-means

"Production-ready" usually means "the approver feels okay." Write it down once, specifically enough that two people reach the same verdict.

## Can you tell when it is broken?

Structured logs. The few metrics that describe user behavior. A health check that fails when the service truly cannot work. A check that returns 200 as long as the process is alive keeps a broken instance in rotation.

## Can you tell when it is degrading?

At least one alert tied to a user-visible symptom, firing before customers notice, with a runbook that someone other than the author has read. If the first sign is a support ticket, it is not ready.

## Can you get out?

The deploy can be reversed, and the reversal has been run at least once. If the release changed the schema, can you go back? Backward-compatible migrations make that possible.

## Does it fail in a contained way?

Every outbound call has an explicit timeout. Retries are bounded and back off. Optional features degrade instead of failing the request.

## Can someone else run it?

More than one person knows how it works. Config is documented and external. No manual step lives only in one person's shell history.

Use the list as shared vocabulary. A service can ship without a rollback path if someone says so and accepts it. The failure is shipping without knowing what's missing.