Dynamic Multi-Cluster Controllers With Controller-runtime
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 15:00· Level 1 | Hall Entrance S10 | Room D·
summary
Introduces a practical extension to controller-runtime that lets controllers reconcile across a dynamic fleet by plugging in a provider concept that discovers clusters (examples: kind provider, KCP provider). Demonstrates a working controller that logs ConfigMaps from multiple clusters, shows switching providers by changing a few lines, and notes a kind provider implementation ≈200 lines. Mentions prototype providers (Cluster API, kind, namespace, KCP), a running demo, and open issues around fair queuing and index management.
AI-written summary · about
chapters (10)
- 0:00Multicluster controllers intro
- 1:52Multicluster reconciliation need
- 3:41Kubernetes controller scaling
- 5:24Multicluster runtime intro
- 9:57How multicluster controllers work
- 14:12Building a multicluster controller
- 20:36Demo of multicluster controller
- 24:39Existing providers and future
- 26:42Community feedback and next steps
- 27:03Q&A
related talks
46mSIG Cluster Lifecycle (Cluster API)
26mWriting Kubernetes Controllers for CRDs: Challenges, Approaches and Solutions
36mThe Multicluster Toolbox
30mBuild Your Own Cluster API Provider the Easy Way
32mCluster API as Code
17mDedicated Cluster management using Cluster API and Kube-Bind
34mMulticluster Ingress Powered by Kubernetes Cluster Registry - Greg Harmon & Nikhil Jindal
42mDeep Dive: Cluster Lifecycle SIG (Cluster API)