Connecting Pool Kubernetes Clusters: Federation and Networks
summary
Describes a concrete multi-cluster Kubernetes networking setup using GKE, VPN/peering and a router reflector to announce pod subnets, combining a global load balancer with per-cluster internal load balancers. Gives measured limits (250 internal LB entries per project, ~1000‑node ingress scale), reports federation adds ~15–25% overhead, and demonstrates a working multi‑region deployment while noting the single reflector reliability risk and practical workarounds for service IP failures across regions.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 4:07Motivation for pooling
- 5:37Multi-region cluster issues
- 8:52Network plugins
- 10:47GKE Internal Load Balancer
- 14:00DNS solutions
- 16:52Federation pros and cons
- 19:47Global Load Balancer and ingress
- 23:20Cloud-neutral solutions and VPN
- 25:50Solution summary
- 27:39Q&A
- 31:16Deployment automation
- 34:48Cloud platform limits
- 38:09Q&A
related talks
46mCluster Federation in Kubernetes: Past, Present and the Future
19mKeynote: Kubernetes Federation
42mSIG Multi-Cluster Deep Dive
37mBuilding the Multi-Cluster L7 Load Balancer
09mMulti-Cluster Networking: A Technology-Agnostic Network Operator for K8 Federation
42mCo-Evolution of Kubernetes and GCP Networking
14mKubernetes Federation for Edge
31mSimplifying multi-clusters in Kubernetes using KubeSlice