English
Menu

What is Canary release?

A release method that first exposes a new version to only some users or requests (for example 5%) and, while watching indicators such as error rate and response time, increases the share step by step if there is no problem. If there is a bug, only some requests are affected, and the old version can be restored quickly. The name comes from canaries taken into coal mines to give early warning of toxic gas. Comparing the old and new versions needs enough requests, and indicator monitoring is often combined with automatic rollback.

Japanese name: カナリアリリース

Canary release appears in 2 practice questions across 1 topic, including Cloud and Container Basics.

Field: Cloud and Container Basics

Try it: questions about Canary release

A service that receives 12,000 requests per hour performs a canary release of a new version, sending only 5% of the requests to it. The new version has a bug that makes 3% of the requests it receives fail; it is noticed after one hour, and all requests are returned to the old version. Which of the following is the number of requests that failed because of the bug during that hour?

  1. A18
  2. B360
  3. C600
Show answer and explanation

Answer: A. 18

The new version received 12,000 × 5% = 600 requests, and 3% of them failed, so 600 × 3% = 18 requests failed. Had all requests been switched to the new version at once, 12,000 × 3% = 360 would have failed. A canary release limits the requests affected by a bug to a small share in this way and increases the share if there is no problem, while watching indicators such as error rate and response time. Comparing the indicators of the new and old versions needs enough requests, so the share and the observation time are decided together.

Canary release in the explanations

Questions that mention Canary release (2)

  1. A service that receives 12,000 requests per hour performs a canary release of a new version, sending only 5% of the requests to it. The new version has a bug that makes 3% of the requests it receives fail; it is noticed after one hour, and all requests are returned to the old version. Which of the following is the number of requests that failed because of the bug during that hour?
  2. Two environments with the same configuration as production are prepared; the current version keeps running in one while the new version is installed and checked in the other, and then the load balancer is switched over to it all at once. If there is a problem, switching back quickly returns to the old version. Which of the following is this release method?

Books on Cloud and Container Basics Ad

As an Amazon Associate I earn from qualifying purchases.

Related terms

Load balancer Blue-green deployment Autoscaling Availability zone Container Container image Docker Immutable infrastructure Infrastructure as code (IaC) Kubernetes

Back to the glossary