What is a canary deployment example?

A canary deployment is a process that tests and continuously monitors your application for potential problems.

Canary deployments are an excellent method to prevent the dreaded "oops, we just found a major bug" situation that can cause problems for your company.

By deploying to the canary environment, you can be certain that everything will go smoothly when it's time for the main rollout. How to do a canary deployment? For this tutorial, we'll assume you have a project called canary-example where you deploy to an environment named canary on the same AWS account as your production and staging environment. First, we'll create a new canary deployment environment in AWS CodeCommit: Open the AWS CodeCommit console In the left navigation pane, select CodeCommit. On the CodeCommit home page, select Create Repository from the menu. Enter a name for your new repository Enter a Description for your repository Enter a region Select the region that you want to deploy your repository to Click Create Repository. Now we'll clone our application to the canary project location in CodeCommit, which will be in the newly created repository. Create a new deployment to your canary environment. First, we need to create a canary deployment. Open the CodeCommit home page Select Deployments from the left navigation pane Click on New deployment from the drop-down menu on the top left Select Canary release from the list of templates. If you see the following message. "This template is currently disabled. Please try again later. This is a temporary error message. If it continues, please contact AWS Support." Select OK to continue

If everything goes well, you should see the following screen: If you don't see the above screen, there are likely some problems. You can check the logs in your repository.

You might get the following error from your deployment process, "A connection was blocked because the data connection was not secure. See inner exception for details. "

In this case, you can ignore the error as the issue won't cause any harm. Now we need to add our canary repository to the deployment environment and configure our application to release to that environment.

How to deploy the code in Azure DevOps?

This chapter demonstrates how to create a new application in Azure DevOps and deploy the code.

For example, we will deploy the Hello Web App. If you want to use other samples such as Angular or ASP.NET Core, you should go to the GitHub repository.

Chapter 2, Working with Build Variants and Build Definitions, explains the concept of build definition and how to create one. However, we don't always want to do everything manually; we can automate tasks such as provisioning, deploying, and configuring as part of the build process. To do this, we will need to deploy a Linux VM where we can do some manual tasks; this is called a build slave. You will then have to execute an application for your environment in a Docker container and upload the images to an Azure Container registry. Lastly, to finish up the deploy, you will need to connect to the Linux VM in Azure using Visual Studio Team Services.

## Setting up the pipeline. For this chapter, we'll use a standard sample project for Azure DevOps.csproj`vsts.user-settings.azuredevops.json`

We should first create a build pipeline that triggers a deployment of the source code after every commit. This build pipeline will be a manual build. In order to set up this build pipeline, the first thing that you will need to do is start a Build and Release pipeline.

What is the difference between ab testing and canary deployment?

To my knowledge, there is no big difference.

In fact, if you read some of the articles on A/B testing and Canary testing it is suggested that they do not differ that much. It's just that one is more about performance (a/b) and one about customer service (canary).

You will end up with 2 apps (a version of app1 and a version of app2) and run them in parallel on your server to see if any data differ. Canary testing works similarly but it is more to catch error and problems at early stage and make a rollback if needed. There are some differences but I think they are minor and it's mostly recommended that you try both techniques and decide which one fits your needs best.