Extending Kubernetes Resource Model (KRM) Beyond Kubernetes Workloads
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 11:00· Level 0 | ICC Capital Hall | Room J·
summary
Argues that KRM can become KRM++ by externalizing APIs into a multi‑tenant control plane (kcp) that splits API provider versus consumer and implements API bindings and API exports so one schema serves many tenant clusters. Covers practical pitfalls (spec vs status, versioning headaches that sparked a 309‑reply thread) and reconciliation patterns (committer pattern, controller runtime). Demo builds a DBaaS: provider runs Postgres CRs, two tenant consumers instantiate identical DB objects via kcp and it succeeds, showing the tree of workspaces model.
AI-written summary · about
chapters (14)
- 0:00Introduction to extending Kubernetes Resource Model
- 1:27Why APIs are challenging to implement
- 3:11Key properties of Kubernetes Resource Model
- 6:03Custom resource pitfalls
- 9:59Evolution and versioning of APIs
- 13:51Reconciliation best practices
- 15:41KRM vs custom solutions
- 18:08Multi-cluster limits
- 19:55KCP for multi-tenancy
- 21:47APIs with KCP workspace
- 23:32KCP as API platform
- 24:15Demo setup
- 24:37Live KCP demo
- 28:07Conclusion
related talks
43mPlatform Engineering Redefined: Evolving Kubernetes Beyond Container Orchestration
06mExtending the Kubernetes API
34mBuilding a Platform Engineering API Layer with KCP
27mBuilding a Platform Engineering API Layer with kcp
40mPanel Discussion: Multi-Tenancy in Kubernetes: Current State and Future Roadmap
37mMulti-Tenancy in Kubernetes: Best Practices Today, and Future Directions
22mTaming the VM Sprawl: KCP Meets KubeVirt
30mExtending Kubernetes Scheduler for Multi-Cluster and Multi-Cloud Workloads