A Brief History of Kubernetes Fleet Controllers & Essential Features
summary
Managing hundreds to thousands of clusters forces a shift to platform engineering with a claimed 100:1 cluster-to-engineer ratio, so fleet controllers must register clusters, propagate workloads, enforce policies, provide observability, lifecycle and failover, and stay cloud-agnostic. Five open source options appear: Clusternet (simple but declining activity), Karmada (scheduling, traffic and recovery), Crossplane (extensible API-driven orchestration), ClusterAPI (powerful but complex), and Rancher: mature, large community with proven adoption.
AI-written summary · about
chapters (16)
- 0:00Kubernetes fleet controllers intro
- 1:46Scaling chaos in clusters
- 3:55Large-scale Kubernetes challenges
- 6:50CNCF & Kubernetes tooling evolution
- 8:14DevOps vs platform engineering
- 12:08What is a fleet controller?
- 14:40Evaluating open-source controllers
- 16:27Analysis of ClusterSet
- 17:53Kubernetes Armada features
- 19:17Crossplane for multicluster
- 20:48Cluster API complexity
- 22:14Rancher fleet management
- 24:43Choosing a fleet controller
- 25:49Mastering fleet management
- 26:55Next steps & alternatives
- 27:58Q&A
related talks
17mKeynote: Extending Kubernetes: Application Orchestration to 1000’s of Clusters in...
03mFrom 5 to 1,300+ Clusters: Declarative Scaling for Private Kubernetes
56mManaging from 1 to 1 million Kubernetes Clusters
24mThe Cluster Is Not the Platform: Operating Industrial Edge Kubernetes as a Fleet
37mKubernetesception: How to run a 1000 control planes
37mManaging and Running Multiple Kubernetes Clusters in Hybrid Setups
19mThe Force Awakens: Managing Millions of Kubernetes Clusters at the Edge
19mKeynote: Kubernetes Federation