[m021478]

I am considering the purchase of an OWC Mercury Rack Pro RAID 0+1 FW800 & USB 2.0 Stripe & Mirror 4 Bay Rack Storage Solutions but I had a question about "Mirroring" in general as a backup/redundancy mechanism...

Mirroring basically creates a mirror image of any data on a drive as it's created (on the physically seperate mirrored drive)...the idea being that if the drive get's corrupted you will always have the mirrorred copy as a backup...

But what happens if it's not the actual drive that gets corrupted, but rather some of the data on the drive...then you would basically have a corrupted copy of the data on both the primary drive and the mirrored drive, correct?

Any suggestions would be greatly appreciated...Thanks!

[tooki]

First a quick standardization of terms: corruption means confusion of or damage to the data on a disk, because software told it to write it that way. If the drive itself messes up, that's a drive failure, which is not considered corruption. Generally speaking, corruption may be repairable, but drive failure never is.

With that laid out, your understanding is fundamentally correct.

Mirroring protects you ONLY against drive failure. It does NOT protect against corruption, because any faulty data written is dutifully written to both disks! So if software messes up and ruins a file or scrambles the disk, or you accidentally trash a file you needed, a mirrored RAID will not help in any way.

Repeat after me: "RAID is not a backup!"

I cannot emphasize that point enough!

So if you are concerned about downtime due to hardware failure, RAID can help. But whether you do RAID or not, you MUST have a separate drive for backup, using backup software (for example, Time Machine), which protects you against the data loss not caused by hardware failure.

[mduell]

Mirroring provides neither backup nor corruption protection; mirroring is only for availability (continuing to work if one drive fails).

For corruption like bit rot, you'll have one good copy and one bad copy and the controller is unable to tell which is good, only that they're different (and if it stripes reads for performance, it won't even be able to tell that). For corruption like an application hosing your file while saving, the mirror will propagate the bad file to all mirrors, immediately.

You need a filesystem with checksumming to guard against the former and backups to guard against the latter.

What are you going to store, how much if it are you going to store, how are you going to access it, and what are you going to connect to it?

Originally Posted by tooki 

First a quick standardization of terms: corruption means confusion of or damage to the data on a disk, because software told it to write it that way. If the drive itself messes up, that's a drive failure, which is not considered corruption.

Are you asserting that bit rot is not corruption?

[tooki]

Correct, I place that under drive failure.

[mduell]

In the context of this thread I think it's worth considering bit rot and drive failure separately, since most RAID1 implementations (particularly at the consumer level) protects against the former but not the latter.

[tooki]

Well, I give my drives NO leeway before I consider them suspect and thus no longer fit for duty. Yes, if a drive silently remaps a few blocks, fine. But if there is anything beyond that, I consider the drive dead, and use it only for non-critical tasks.

[mduell]

How do you detect the 1 in 10^14 bit read error, to know that hard drive is no longer reliable?

[tooki]

If the OS gives me errors that a file cannot be read, and even repeated attempts to read it fail.

[mduell]

Are you using ZFS mirroring on OpenSolaris or something? OS X/HFS+ isn't going to detect the majority of flipped bits.

[tooki]

No, of course not! Have you never gotten a Finder error trying to open or copy a file?!?

[mduell]

Certainly, but that's not indicative of or even necessarily related to bit rot/random bit read errors.

I wonder if the OP will ever come back to this thread.