Blue/Green deployment strategy

#loadbalancing   #cloud   #docker   #aws   #nginx  
Canary deployments

Canary deployments

Canary deployment strategy also known as blue/green deployment, is an industry standard protocol for deployment with zero downtime by utilizing two separate environments and switching during source code updates or rollbacks. Canary deployments have a lot of advantages when it comes to multi-server configuration and aids in technical cases like version rollbacks, and a/b testing.

What is blue-green deployment?

The strategy gets its name from the idea of having two concurrent environments and switching between environments. The environments have different code revision signatures.

The blue environment is the old environment.
The green environment is the newer and more updated environment.

Since the deployment process changes the web pointers from the blue environment to the green environment, it is known as the Blue-Green deployment strategy.

Of the many benefits provided by the blue-green deployment strategy, the most important few are:

1. Software & service downtime reduction

If production deployments relied on a single environment, there would definitely be server downtime, owing to the time cost of source code updates and patches. However, if there are two environments running concurrently, this downtime gets reduced to nil given there occurs no mid-modification issues.

2. Experimental feature testing

When new features need to tested publicly, this strategy offers significant advantages. Since the idea of this strategy involves having 2 environments, a small portion of the users can be exposed to one of the environments with the experimental features. Depending on the response of the users, devs can take further decisions about the impact of the new features, change the ratio of exposure or totally upgrade to the new version.

3. Isolation for testing and production environments

One of the better uses of the blue-green deployment strategy is keeping the development and production environments separated from each other. Devs should never have to worry about writing new code or changing infrastructure which will directly impact the live servers. This will only cause public frustration.

Something which can easily turn into memes.

4. Crashes and rollbacks

One of the more common and most frightening term for any software company on earth would be crash. But even though, code has been tested hundreds of times before release, there might be times when a few bugs may jump out of the bucket. Critical ones. Having a blue-green deployment technique can be a life-saver during these cases, as teams can switch back to the stable older environment where those bugs are not encountered, thus giving them freedom and salvation from the torment of the public and media as they sit down to fix the bugs and redeploy again. If some different deployment technique (in-place deployment) was used instead, rollbacks would become very long as there would be no older environment to switch back to.

In the industry of cloud and web tech, blue-green or canary deployment are still one of the main and most popular deployment strategies that is used to upgrade running servers with the latest production code. Every big software company (Facebook, Google, Amazon, and so on..) has used or is still using this strategy for their deployments because of the conveniences it provides.

However, a downside to this strategy is that, as the switching of the environments are handled by the load-balancers, a switching time-delay due to DNS propagation occurs. DNS servers are updated with the correct server DNS names(their unique address on the web) via propagation, which means, it takes time to reach every domain name server across the planet. As a result, people of some regions of the world may not be able to access the changes in the servers as soon as the deployment happens, because the DNS servers closest to them are still pointing to the older servers. Sometimes, when a new deployment has taken place with bugs in the code, a rollback although issued instantly will not be perceived in every part of the world. So people will still be looking at the buggy version of the software even though the buggy version has been correctly rollbacked from.

Although these can occur, proper trial and testing can help resolve these problems before they are released to the open world. Compared to the bigger pool of advantages, these problems are easily overlooked and are the reason why blue-green deployment strategies are still preferred today.

Avatar

Sowmen Rahman

Junior Software Engineer, Arcade Studios

@sowmenrahman
CS senior undergraduate @CUET, and ML hobbyist. Enthusiastic about code automation and clean architecture. Full stack native app and game developer.
10

Authority

174

Total Hits

Discussed tools
NGINXDocker ComposeAmazon ELBDocker