Rook-Ceph storage for Kubernetes
Persistent block and shared volumes for Kubernetes, served by the Proxmox Ceph cluster through Rook in external mode.
Context
The Kubernetes platform needed persistent storage for stateful workloads. A Ceph cluster was already running on the Proxmox nodes.
What I did
- Rook in external mode. Instead of running a second Ceph cluster inside Kubernetes, Rook connects its CSI drivers to the existing Proxmox Ceph cluster.
- Volumes. RBD for block volumes and CephFS for shared volumes.
- Tuning. Adjusted Ceph placement groups.
- Validation. Checked throughput across the 25 Gbps replication links.
Outcome
- Stateful workloads get replicated block and shared volumes that survive a node failure.
- Throughput validated across the 25 Gbps replication links.