I do something similar with aufs and snapraid. Aufs is responsible for the storage pool (union mount). I look forward to replace it with something integrated i

I do something similar with aufs and snapraid. Aufs is responsible for the storage pool (union mount). I look forward to replace it with something integrated in the Linux kernel (e.g overlayfs). Snapraid is responsible for the redundancy. It kind of implements an offline RAID5 scenario per file —instead of per disk block—, it only maintains a parity disk. Of course since it is offline, you have to update the parity every time you change a file. Usually once per day or per week is ok for long term storage. This approach offers many advantages. Every data disk just works everywhere since it has a common filesystem in it. Via parity data you can undelete files, find duplicate files and correct read errors. You won't lose all drives if you lose more than parity+1 drives. It allows scrubbing and rebuilding.