Overview
"Blue-Green deployment" is an application release model, software deployment strategy and deployment technique, which offers:
Blue/green deployments can mitigate common risks associated with deploying software, such as downtime and rollback capability
- Amazon in the AWS White paper on blue/green deployments
How does "Blue-Green Deployments" work?
In Blue-Green Deployments, we have 2 servers, named green and blue. Both of them have identical environments but different versions. In the blue server, you have your production application which handles almost all of your traffic. Whereas in the green server, you have, a semi-production or staging version.
Now we test the green server if it works and is ready. We start slowly channelling away traffic to the green server. In other words, we follow a rolling release to publish the green server version.
So for instance, we have blue running on v1.0.0 (prod), and green on v1.0.0-beta (staging). We test the green server's version internally, then roll out the new version in chunks to our users. In this example, we give 25% of the people access to use it.
And this is repeated till eternity ✨ Here is the graph of the same scenario:
What happens if my beta app breaks?
Let's take a second scenario, in this scenario, somehow the application passes the internal testing phase, but when users use it, it breaks apart 🥲. We can do an instant roll-back by redirecting all the traffic to the blue server!
Rolling releases with Blue-Green Deployments
Since we have 2 servers where we are diving the traffic, we get to offer a certain feature to a set of users. This provides different benefits:
You minimise risks since your release does not affect your entire user base
A/B Testing allows you to analyse user groups' performance (both system & conversion).
Feedback Loops: You can constantly collect feedback from users and iterate to improve your application.
You can send out the version to your users by directing traffic based on various factors such as:
Randomly: You can send it out randomly to your users
Geographically: Want a feature to only be available in the USA while you test it? You can!
Beta-testers: Enrol users in a beta program or a select group of users like your friends and team 👀
So basically it is like feature-flags but for versions!
More benefits
Production testing! Since the app is on a replica of the production environment you can say we're testing on production 💥
Since you are switching servers and resources, you do not have to worry about upgrading the application without breaking everything else!
Since with the cloud, we can tear apart and provision new services instantly, it is also more lucrative/cost-efficient for us.
Challenges
It does not make sense to have 2 databases and to sync between them for the environments, instead, we do the obvious and use the same database. The problem occurs when you make a breaking change to your schema because then your production version falls apart as it no longer can query for the same data schemas.
Other challenges include the inability to use the components of this method due to project, regulation or infrastructure-based restrictions.
Further resources
