Building for the (Inevitable) Next Cloud Outage
summary
Three k3s Kubernetes clusters in New York, Amsterdam and Sydney advertise the same BGP anycast address via the Bird daemon so user requests hit the nearest cluster; when one cluster stops announcing its route traffic converges to the next site in under a second. Workloads use NATS JetStream event sourcing so CRUD events replicate across clusters and materialized views reach consistency in milliseconds–seconds; long‑lived connections must reconnect after failover. The demo runs end‑to‑end and the system continues serving with near‑zero visible downtime.
AI-written summary · about
chapters (4)
related talks
31mAn Adventure in Multi Cloud with Kubernetes and Distributed Data
41mAn Adventure in Multi-Cloud with Kubernetes
23mBuilding Unified Global Load Balancing for the Edge With k8gb
13mBuilding Reliable Infrastructure for LLM Workloads on K8s
35mHybrid Cloud Powered by Kubernetes [I]
34mCloudy With a Chance of Kubernetes: Going From One To Three Cloud Providers
33mKubernetes-Driven Multi-Cloud Deployments: Maximizing Scalability and Flexibility
33mWhat We’ve Learned Building a Multi-Region DBaaS on Kubernetes