witty-musician-14961
11/06/2026, 7:57 AM<http://rook-ceph.rbd.csi.ceph.com|rook-ceph.rbd.csi.ceph.com> as a validated driver. I also see the KB on Rook with an existing external Ceph cluster. What I'd like to know:
1. Upstream ceph-csi vs Rook external mode: Is installing upstream ceph-csi directly (Helm chart, provisioner <http://rbd.csi.ceph.com|rbd.csi.ceph.com>) supported or tested? Or is the recommended path Rook in external-cluster mode, pointed at our cephadm cluster, so the driver is <http://rook-ceph.rbd.csi.ceph.com|rook-ceph.rbd.csi.ceph.com>?
2. Feature parity: With either option, do VM images, root volumes, live migration and snapshots behave the same as in the validated Rook RBD matrix? Are there CDI StorageProfile or csi-driver-config settings we should know about, like clone strategy or volume mode Block?
3. Host prereqs: Are the rbd module and the /var/lib/ceph persistent paths still needed if we use ceph-csi without Rook?
4. Production use: Is anyone running a non-Rook external Ceph in production with Harvester? Any gotchas with upgrades, Ceph version compatibility, or multi-cluster setups sharing one Ceph?
Thanks!