Skip to content
ML Academy

Introduced on day 19

Canary release

A partial, time-limited deployment of a change to a small share of traffic, evaluated against the rest as a control, with stop criteria agreed in advance.

In detail

The canary should be large and long enough to be representative, use a small set of attributable metrics, and be compared with the control rather than with a before period. A bad canary is rolled back, ideally by one step such as re-pointing a registry alias.