Shipping at the Edge: Migrating a Coffee Subscription Platform to Cloudflare Workers

The article describes migrating the Brewly Store coffee subscription platform from a traditional AWS EKS setup to an edge-first architecture using Cloudflare Workers and the Hono framework. The migration aimed to reduce latency, improve scalability for subscription spikes, and enhance developer experience, while maintaining security through GPG-signed commits and GitOps-driven deployments with ArgoCD. The author notes that the primary challenge was shifting from a server-centric mindset to managing distributed serverless functions, requiring improved observability.

Building a D2C platform isn't just about the product; it's about the resilience of the delivery chain. I've been working on Brewly Store, a coffee subscription service, and recently decided to tear down our traditional setup to embrace a more modern, edge-first architecture. The Tech Stack 🛠: Runtime: Cloudflare Workers using the Hono framework . Language: TypeScript. Infrastructure: AWS EKS for core services. Delivery: A full GitOps pipeline using ArgoCD and GitHub Actions. Why the Migration? We moved from a containerized AWS ECR setup to Cloudflare Workers. Why? Latency: Moving logic closer to the user. Scalability: Handling subscription spikes without managing scaling groups. Developer Experience: Hono provides a lightweight, expressive way to handle routing that feels incredibly fast. The Security-First Mindset 🔒: As a DevOps specialist, I believe infrastructure is only as good as its security. For Brewly, I’ve implemented: GPG-signed commits to ensure the integrity of our code. Strict secret management using tools like Kleopatra and PwPush. GitOps-driven deployments where ArgoCD ensures the cluster state matches our repository exactly. Lessons Learned from the Trenches: The biggest challenge wasn't the code—it was the shift in thinking from "servers" to "distributed functions." Debugging edge cases in a serverless environment requires a robust observability stack, which we are currently refining within our EKS cluster.