I'm building some new infrastructure in AWS and plan on having separate accounts for dev, stage, and prod environments. I'd like to implement a blue/green deployment strategy at least for production, and I've seen a few different ways this could be done, an obvious one being CodeDeploy's built-in blue/green deployment capability.
Someone suggested to me that the stage/prod environments could be thought of as the blue and green environments. They are, after all, supposed to be identical environments, but I hadn't thought of stage being something that could act as a production platform. I think stage is supposed to be for end-to-end testing, final verifications by QA, demos, etc.
So, I believe stage should continue to be for internal purposes only. But we may need two production accounts. At any given time each of them would be expected to contain a combination of inactive and active services (blues and greens). DNS (Route 53) would be in a shared account, and when a release occurs it would switch routings from resources in one account to resources in the other one.
Then, it kind of seems like having stage is less useful when you have blue/green. When a new release is in the inactive state and going through the most comprehensive tests, it's essentially what stage does. And we don't even really have QA--developers and our customers are the QA. So can we get rid of it? I have heard praises sung of having a straight-to-production strategy whenever automated testing is taken seriously enough.
Is this sensible?