Canary analysis is often used in conjunction with continuous deployment to ensure that changes are tested in a controlled manner before being rolled out to production. The term "canary analysis" comes from the practice of using canaries in coal mines to detect dangerous levels of gas. Just as canaries were used to alert miners to potential danger, canary analysis is used to alert system administrators to potential problems or changes in the behavior of a system or service.
In canary analysis, a baseline of normal behavior is established for a system or service, and this baseline is used to compare against subsequent measurements. If the system or service exhibits behavior that deviates significantly from the baseline, this may indicate a problem or change in the system that needs to be investigated.