Hi all
we have several customers running openshift on openstack.
If you follow the platform guides, the openshift products integrate best with OCS (gluster), while the openstack engineers prefer ceph as cinder backend.
Hence when following the default guidelines you end up with gluster on ceph, having duplicate replication (on gluster level and on ceph level), making the storage for the customer costly. (3way gluster rep x 3way ceph repl = 9 times replication of the data!
What are the alternatives to get rid of this waste of storage?
i see:
1/ cephFS as PVC on openshift using cinder apis
2/ separate gluster on BM next to openstack
3/ 3 way gluster rep x 1 way repl ceph + making sure each gluster brick ends op on a different ceph host?
4/ others?
many thanks for any reference to a doc, person of the storage team of experiences from the field
filip
