Scaling Multi-cluster Kubernetes at Teleport Cloud
summary
Runs a dedicated Teleport per customer across six regions, handling over 10,000 pods and 100,000 reverse tunnels; migrated from gravity to EKS and uses Route 53 latency records + regional NLBs for stable ingress. Uses Envoy with ALPN (via an envoy Gateway fork) and a custom service-label flipper plus minReadySeconds zero-downtime to roll proxies without breaking long-lived tunnels. Deploys coordinated rollouts with an open-source sync controller (Apache‑2) and uses psyllium Global Services for cross-region pod networking; architecture is live and production-proven.
AI-written summary · about
chapters (15)
- 0:00Intro
- 0:27what is teleport
- 1:25teleport's architecture
- 2:31teleport cloud
- 4:02needs: ingress, deployment, container networking
- 5:08anycast, nginx (oss)
- 7:32ingress
- 11:08zero-downtime
- 13:21needs (ingress)
- 14:28deployment
- 15:43cross-cluster reconcilers
- 22:03container networking
- 24:14user journey
- 24:24open source
- 25:15thanks!
related talks
45mMulti-Kubernetes, Multi-Region — Help!
35mKubernetes journey
22mSay Goodbye to Manual Kubernetes User Access Onboarding
47mFast and Furious Multi-Cloud Connectivity
34mLow Latency Multi-cluster Kubernetes Networking in AWS
33mKubernetes-Driven Multi-Cloud Deployments: Maximizing Scalability and Flexibility
15mIdentity-native infrastructure access on AWS with Teleport
27mKubernetes Everywhere: Lessons Learned From Going Multi-Cloud