Don't Start With 500 Clusters: Patterns in Scaling Multi-Cluster Kubernetes…
summary
Two companies adopt Cluster API to automate fleet management and avoid manual toil: one migrates 60 clusters in 3 months and shrinks provisioning from a week and a half to under six hours, the other gets bare‑bones clusters ready in ~15 minutes for multi‑cloud and BYOC deployments. Both build management clusters with Argo CD and custom companion operators, hit Azure provider bugs that caused node loss, and achieve automation-before-scale, provision bare‑bones clusters in ~15 minutes, reduced provisioning from 1.5 weeks to <6 hours.
AI-written summary · about
chapters (7)
- 0:00Multi-cluster Kubernetes scaling intro
- 3:54Avoid scaling to 500 clusters
- 8:46Atoms.co’s cluster API adoption
- 13:20Ditto’s multi-cloud challenges
- 24:40Fleet management pattern
- 26:45Key learnings from scaling
- 27:26Closing remarks & Q&A
related talks
47mPostmortem: A Modern Approach to Managing Kubernetes Clusters
36mMulti-Cluster Kubernetes Quick Delivery with IaC and GitOps
39mCluster API & Cluster Autoscaler
45mLarge-scale kubernetes deployment, which tools to use and how - Rachid Zarouali
21mPlatform Perseverance: Taming 1,000 Kubernetes Clusters
30mScaling Kubernetes to Thousands of Nodes Across Multiple Clusters, Calmly
31mHow to Migrate 700 Kubernetes Clusters to Cluster API with Zero D...
29mNavigating the Multitenancy Maze: Avoiding Kubernetes Antipattern Pitfalls