Description of problem: When ovirt-node is installed, there is no way to set up a mirrored root disk. In production, if the disk on the node fails, the node will go down. Assuming high availability, and central storage for VMs and their images, the engine will bring the VMs up on another node in the cluster. However, there would be no downtime, and no moving of VMs at all if the node survived a disk outage. MD is already built into the Linux kernel on the node which should make this possible.
That sounds interesting. Assuming that there are enough PVs this might be doable.
Interesting, but i'd like to lump this into the investigation of using the anaconda storage methodology (blivet) and the re-write of our storage management.
We're considering a different node implementation for oVirt 4.0. Hence, closing this RFE. Please re-open if relevant.