Which of the following describes a Canary deployment?

Which of the following describes a Canary deployment?

• A. Gradually route increasing fractions of traffic to the new version; monitor latency, error rate, and saturation; proceed if metrics stay within thresholds.
• B. Deploy to all users at once; monitor for issues; rollback only if critical failures occur.
• C. Use a feature flag to toggle release; no traffic routing changes.
• D. Release is manual; no automated metrics used.

Answer: (A)

Canary deployment is a staged rollout strategy. You start by directing a small portion of live traffic to the new version and watch how it behaves in production. You track key signals such as latency, error rate, and saturation, comparing them to predefined thresholds. If the metrics stay within acceptable bounds, you gradually shift more traffic to the new version, continuing this ramp-up until the full population is on the new release. If issues appear, you can halt the rollout or roll back to the previous version to limit impact. This approach reduces risk by catching problems early with limited exposure before a full deployment. Deploying to all users at once is a full rollout and lacks the early feedback window Canary provides. Using a feature flag to toggle a release without changing traffic routing doesn’t test the new version in production with progressive exposure. A manual release with no automated metrics also misses the automated risk controls Canary relies on.

Canary deployment is a staged rollout strategy. You start by directing a small portion of live traffic to the new version and watch how it behaves in production. You track key signals such as latency, error rate, and saturation, comparing them to predefined thresholds. If the metrics stay within acceptable bounds, you gradually shift more traffic to the new version, continuing this ramp-up until the full population is on the new release. If issues appear, you can halt the rollout or roll back to the previous version to limit impact. This approach reduces risk by catching problems early with limited exposure before a full deployment.

Deploying to all users at once is a full rollout and lacks the early feedback window Canary provides. Using a feature flag to toggle a release without changing traffic routing doesn’t test the new version in production with progressive exposure. A manual release with no automated metrics also misses the automated risk controls Canary relies on.