^misantropia^ wrote:
Not even that is entirely safe. If you are doing hardware RAID you'll need identical drives (up to the firmware version, often) so they will have the same MTBF. If I got a nickel for all the times drives died in a cluster almost simultaneously, I'd have... well, a few nickles at the very least.
It depends on how much you value your data, I suppose. Losing a bunch of MP3s is a nuisance but no more. Your company's financial records less so.
That's not true. I've had to replace drives 3 times in servers in the last week and a half, and in two of those cases I replaced them with drives that were from a completely different manufacturer. To any decent modern RAID controller, all that matters is that the disks are the same size and have the same performance specs.
Also, your example of multiple drives dying in a cluster, while valid, doesn't apply so much here. The odds of having multiple drives die in a system increases with each additional drive you have in it. If you have a cluster/storage group with 50 physical disks in it, your odds of having more than one drive fail are exponentially higher than if you had a system with two disks.
The mirrored + spare system vastly decreases your odds of having a non-recoverable failure, for a few reasons. First off, there are only two disks, so your odds of having two fail at the same time are low -- but even if you DO have two disks fail at the same time, it doesn't matter, because you still have a perfectly good, bootable drive sitting on your shelf that has everything on it except what you've done in the last few weeks.
Second, since you are deciding when to pull a drive for backup purposes, that means that you will only be pulling your backup at a time when you know your system is good and your data is intact. That is the upside to not being automated. If you use an automated/scheduled backup system, you have a very real chance of backing up corrupted data, and therefore not being able to recover it.
The only real disadvantage to the mirror + spare system is that you are wasting a whole disk and also that, in order for it to work, you have to be able to fit all your data onto a single disk (unless you want to buy a bunch of disks and get really complex, like a cycling mirrored set of stripes for instance).
Since the disk is on a shelf, you don't have to worry about corruption, power surges, etc. It's a good system that doesn't take much work to setup. Just make sure you shut down the machine before you pull out your spare drive (even if it's hot-swappable), so you don't get any corrupted files from interrupting disk activity.