Zero-Downtime migration from ingress NGINX to Envoy Gateway

A case study of migrating from Ingress NGINX to Envoy Gateway on AWS with zero downtime. The team evaluated Gateway API controllers using CNCF alignment, mTLS support, and annotation parity as filters, ultimately selecting Envoy Gateway. Their first migration attempt caused a brief outage due to DNS TTL lag. The solution was weighted DNS records via ExternalDNS and AWS Route 53, running both Ingress and HTTPRoute simultaneously with weight 100/0, then swapping weights to shift traffic without deleting DNS records. This approach also makes rollback trivial. The post also covers production challenges like multi-namespace hostname management and the upcoming Gateway API 1.5 ListenerSet feature that addresses separation of concerns between infrastructure and application teams.