Blue-Green vs. Rolling Updates for Zero-Downtime Migration via Travis CI

When migrating a small application using Travis CI, the goal is to achieve zero downtime while managing infrastructure costs and deployment risk. Because Travis CI acts as an orchestration engine rather than a traffic manager, the strategy for shifting traffic must be handled via external provider APIs or orchestration platforms.

A Blue-Green approach requires deploying a full parallel environment, doubling the infrastructure capacity during the transition. Conversely, Rolling updates leverage platforms like Kubernetes or AWS ECS to replace instances incrementally, avoiding the cost of a duplicate environment but introducing a period where two different versions of the application coexist.

A specific concern arises regarding stateful connections and long-running processes, as Travis CI cannot natively track or drain active sessions during the cut-over.

• Which approach better mitigates the risk of version mismatch for stateful sessions?
• How should the trade-off between infrastructure cost and deployment safety be weighed when implementing custom health-check scripts in Travis CI?