jamesiarmes
commented
8 years ago 
I wrote a tool called howiroll to do ASG updates, because I couldn't find documentation in CloudFormation to support that it handled ELB connection draining properly. In that tool, I noticed I didn't add new instances before removing old ones. I'd like to port that functionality into Moonshot and improve on the methodology for rolling out launch configuration updates.
That being said, you're asking about CodeDeploy. I suppose it would be possible to modify the CodeDeploy DeploymentMechanism with knowledge of modifying and waiting for ASG capacity to change, but I think a good first-pass solution is simply to always scale your ASGs for N+1 (or greater) capacity. In critical infrastructure, probably N*2+1. Otherwise, how are you going to sustain an Availability Zone failure?
I'm not saying I'm against this feature, just that I think it has a low priority because it can be mitigated in production by throwing money at the problem, which is generally cheaper than throwing engineering time at a problem.
Not sure if this is within the scope of this project, however the goal of this enhancement is to maintain at least the original number of healthy hosts that you started with throughout the deployment. In an anti-fragile system, it should strengthen during risky operations like code deploys.
For example, my application has 3 instances. During deployments, this temporarily dips to 2 healthy hosts since each instance has to stop the service, deploy the code, and then restart the service. See the graph below:
When performing AMI updates, CloudFormation temporarily increases the desired capacity during the rolling upgrade to maintain the number of healthy hosts throughout the process. It would be great to mimic this behavior, whether that is core to the Moonshot tool or some sort of opt-in thing with via some extension.