Duolingo's Kubernetes Journey: From ECS to EKS

Duolingo's Kubernetes Leap: A Deep Dive

Duolingo's migration from AWS ECS to EKS, as presented by Franka Passing, offers a compelling case study for organizations facing similar scaling challenges. The decision to move to Kubernetes is well-justified by the need for a more feature-rich, open-source ecosystem and multi-cloud flexibility, even while acknowledging ECS's prior suitability. The emphasis on GitOps with Argo CD for improved deployment strategies like blue-green deployments and ephemeral environments is a critical takeaway. The 'cellular architecture' for environment isolation is particularly noteworthy for its ability to de-risk platform changes and manage dependencies across development and production. The 'reports from the trenches' on managing developer trust, AWS rate limits, and productionizing early adopters provide a realistic perspective on the complexities of such large-scale migrations. The structured approach, dividing the project into building the foundation, productionizing early adopters, and general adoption, is a sound project management strategy.

However, the presentation, while rich in technical detail, could benefit from more explicit discussion on the long-term operational overhead of managing a Kubernetes cluster compared to ECS, beyond the initial investment. While GitOps and automation are highlighted, the ongoing maintenance of the Kubernetes control plane itself, security patching, and the potential for escalating complexity as more teams adopt the platform are areas that might require further elaboration. The mention of AWS rate limits is a practical concern, but the specific strategies Duolingo employed to navigate these could offer more actionable advice. The 'VIP support' model is a good approach for onboarding, but its scalability and potential for creating bottlenecks in the future could be a point of consideration for other organizations looking to replicate this success. The IPv6-only pods mention is intriguing but lacks depth in the provided summary, leaving room for further technical exploration.

Key Points

• Duolingo is migrating over 500 backend services from AWS ECS to EKS (Kubernetes).
• The migration is driven by the need for Kubernetes' richer feature set, open-source ecosystem, and multi-cloud support.
• GitOps with Argo CD is central to the migration, enabling advanced deployment strategies like blue-green deployments and ephemeral environments.
• A 'cellular architecture' is employed for environment isolation, allowing for safer platform changes and testing.
• Key challenges include managing developer trust, navigating AWS rate limits, and productionizing early adopter services.
• The project is phased: building the foundation, productionizing early adopters, and general adoption.

---

📖 Source: Presentation: Duolingo's Kubernetes Leap