Making Ceph Awesome on Kubernetes with Rook
summary
Rook runs Ceph as a Kubernetes-native operator controller, declaring desired state with CRDs and continuously reconciling to create and heal Mons and OSDs, MDS for filesystems and RGW for object stores. It keeps no external state, lives in the control path while clients talk directly to Ceph, and uses flexvolume -> CSI for mounts. Networking overlays can impact performance but host networking is an option. Project is pre-1.0 with tens of production deployments reported and ongoing stabilization toward production readiness.
AI-written summary · about
chapters (12)
- 0:00Introduction to Rook
- 1:07Kubernetes as cluster OS
- 2:38Storage for stateful apps
- 4:25Why Rook for Ceph
- 5:07What is Rook
- 6:56Rook in action
- 8:40Rook as Kubernetes operator
- 11:08Rook architecture
- 12:36Division of labor
- 14:22Converged Ceph clusters
- 15:48Future of Ceph on Kubernetes
- 16:55Q&A session
related talks
12mRook: Why Would You Ever Deploy Ceph Inside Kubernetes?
41mRook - Running Ceph Using Kubernetes
26mRook: Intro And Deep Dive With Ceph Storage
55mIntroducing Rook
35mDeploy Ceph in Kubernetes: Rook Deep Dive
29mSimplifying Storage in Kubernetes with Rook-Ceph
30mRook: Intro and Ceph Deep Dive
31mRook: Intro and Deep Dive With Ceph Storage