Local linux machine with three 2T drives (WD Red's), setup in RAID 5 for redundancy against a drive failing. LVM running on top of the RAID 5 drive trio to all

Local linux machine with three 2T drives (WD Red's), setup in RAID 5 for redundancy against a drive failing. LVM running on top of the RAID 5 drive trio to allow slice-and-dice allocation of the space. The LVM backup partition only consumes 220G. rsnapshot (http://www.rsnapshot.org/ http://www.rsnapshot.org/) from other machines (they are Linux as well) onto the Linux backup server (so now there are two physically separate copies, master machine and backup machine). And rsnapshot provides "point in time" views of the backups. Periodically rsync the rsnapshot tree to an offsite location with enough space to hold the 220G worth of backup data. What I don't do, which you imply you did with your Drobo, is have the only copy of anything truly important on the RAID5 array only. If Linux md or LVM corrupts the backup partition I just remake it, and start rsync'ing again (and pull old snapshots from the offsite). If one drive fails in the RAID5 array, I'll replace the drive and rebuild the array, but the data should still be there. If two drives fail in the RAID5 array close on each others heels (so that the whole RAID array goes away) then this is the same as if md or LVM corrupted the data. "Backup" does not mean having "redundant protection" in one device, it means having plural separate, fully independent, copies such that corruption/loss of one copy is meaningless to the other copies.