A team sends a new build to two percent of live users, watches error rates and latency for an hour, and then raises the share in steps only if those metrics stay healthy. What is this rollout pattern called?
Show answer and explanation
Answer: D
This is a canary release. The new version receives a small slice of real traffic first, and the team decides whether to continue based on observed health metrics. Each step increases exposure only after the signals remain acceptable, which limits the blast radius of a bad release. A recreate strategy stops the old version before starting the new one, a rolling replacement swaps instances gradually without a metric-gated traffic split, and blue/green switches all traffic in one move.
Why the other options are wrong
- A. A recreate deployment shuts down the existing version completely and then starts the new one. It causes downtime and does not expose a small share of live users to the new build first.
- B. Rolling replacement gradually swaps instances for new ones, but it does not usually split traffic by percentage based on metric gates. The scenario describes a controlled exposure step, which is what distinguishes the canary pattern.
- C. Blue/green keeps two full environments and switches all traffic at once. The scenario gradually raises a small share of traffic, so it describes a canary rather than a single cutover.