I'm working on a project with AWS CDK in Python where I've set up a load balancer, security groups, and an auto-scaling group (ASG). This will act as a central pipeline, and I'm currently triggering deployments via GitLab. I need suggestions on how to implement a blue-green deployment strategy.
Here's the scenario: I already have an ASG named **auto-scaling-group-7ea57ea1** based on a Git commit SHA, which provisions one or more EC2 instances.
For each new deployment by a team, I want to build a fresh ASG, let's call it **auto-scaling-group-9ff0d223**. This ASG will provision new EC2 instances.
1. If the health checks on the new EC2 instances are successful, I'd like to deregister the instances from the old ASG (**auto-scaling-group-7ea57ea1**) from the load balancer. It's important that I do not terminate the old instances. The new ASG must be registered to the target group instead.
2. If there are issues with the new deployment (like bugs causing wrong results), developers should be able to revert back to the **auto-scaling-group-7ea57ea1** without losing any instances.
Currently, my code supports rolling deployments, where the old ASG is terminated upon successful health checks of the new ASG. Any help or ideas on how to achieve this blue-green deployment pattern in AWS CDK would be greatly appreciated!
3 Answers
It sounds like you're looking for a blue-green deployment setup. You should definitely check out using blue/green deployments within CDK, particularly through ECS deployment groups. They can really help manage environments without deleting your existing ASGs right away!
Just to clarify, you mentioned ECS but I'm more interested in using EC2. Can these strategies still apply?
I'm curious about your ultimate goal. Why are you explicitly naming your resources like this? It seems like it might complicate things more than necessary.
Good question! I'm designing a common pipeline meant for multiple applications across different teams. I'd love to know how you set yours up.
You might want to look into setting up two separate auto-scaling groups as a part of your deployment. AWS has a detailed document on how to retain metrics across blue-green deployments which could be useful for your situation.
I totally agree! The concept of blue-green deployment seems perfect for your needs. I’m in a similar boat and found it super helpful.