Blue-Green vs Rolling: How Engineers Decide Who Deploys and When

A developer dialogue published on DEV Community explores the real-world tradeoffs of container deployment strategies, framed as a conversation between a junior and senior engineer. The discussion begins with the simplest approach—stopping the old container before starting the new one—and explains why this 'recreate' strategy causes downtime that can be costly for production services. The senior engineer then introduces blue-green deployment, where two full environments run in parallel and traffic is switched instantly via a router flip, enabling near-instant rollback without redeployment. The conversation highlights that blue-green requires double the infrastructure, making it expensive for smaller teams, which is why rolling deployments—replacing instances gradually behind a load balancer—are more commonly used in practice. The core engineering question posed is not how to run a new version, but whether old and new versions can safely overlap long enough for a zero-downtime cutover.