Redundant Datastore for HA


We are using OpenNebula currently on a single server without any HA. On this Server is a NFS share of the datastore directory exportet and mounted on all of our Hypervisors.

Now we would like to use the HA feature of sunstone and the oned, but how can we make the datastore redundant in case our first server (currently the only with the datastore on it) went down?

I have the following Idea:

I plan to use Syncthing to sync the datastore directory from the main existing server to all HA-Server in Cluster (3 Server) and sync them all together to make this directory redundant.
Since I’m using the floating IP on all HA-Servers I would change the NFS-Mount on all of our Hypervisors to the floating IP.

Is there anything wrong with my idea? Do you have any suggestions?

We are currently using 350+ VMs, it is important for us that there will be no downtime.

We also use Ceph for our Data, is it possible to change the NFS-Shared Datastore to a CephFS without loosing all the existing VM Data? We are not able to recreate all the VMs beacuse they are used in a productive environment with many customers.

I would be very thankful for any input on this.

Best regards