Multicluster Kubernetes, today and tomorrow
summary
Defines a cluster set as clusters with shared trust and names and outlines four concrete pain points: config drift, scheduling/workload placement, multi-cluster storage/stateful migrations, and networking/service discovery. Recommends the about API (cluster-scoped CRD for cluster properties), the MCS pattern of ServiceExport/ServiceImport for service discovery, points to Work API work for placement, shows GKE’s Gateway API multi-cluster classes for ingress load balancing, and highlights KEP-3335/statefulset-slices for migrating stateful workloads.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:37Applications are multicluster
- 2:40What is multicluster
- 8:15Who can be in the cluster set
- 10:25Reasonable degree of trust
- 12:52Cluster Admins
- 14:47Problems
- 19:27MCS API
- 22:33Trust
- 25:50When not to use multicluster
- 31:36Split brain problem
related talks
06mKeynote: From One to Many, the Road to Multicluster
23mSIG-Multicluster Intro And Deep Dive
18mSIG Multicluster Intro
50mMulti-Cluster Kubernetes and Service Mesh Patterns
31mHere Be Services: Beyond the Cluster Boundary with Multicluster Servi...
34mSIG Multi-Cluster Intro – Christian Bell, Madhu C.S., & Nikhil Jindal, Google
37mIntro: Kubernetes SIG Multicluster
36mThe Multicluster Toolbox
