Can a single unencrypted drive removed from a RaidZ2 pool be read?

[nanoduckie]

I have a WD Red 4TB drive that is about to fail and I'm going to RMA it. However I am a little worried as my drives are unencrypted and I don't want strangers to be peeking into my files. Could data from such a drive be read by a 3rd party?

 

[nojohnny101]

Just pull it from the server and zero it out with some software (there are dozens to choose from).

 

[Stux]

You can actually wipe it (and zero it) from the FreeNAS GUI.

Offline it, then wipe it.

FWIW, there would only be fragments of data in a broken non-trivial file system.

Someone would have to be dedicated to get anything.

 

[nanoduckie]

This means that if I do not wipe the drive, the data can be read?

 

[Stux]

This means that if I do not wipe the drive, the data can be read?

Yes, if you do not zero the drive, then there is data on it, and it can be read. maybe not easily, but if you don't wipe the drive, there is data on it.

But you can't just pull off files and pictures off the drive. There are only fragments of files on the disk.

 

[nanoduckie]

Yes, if you do not zero the drive, then there is data on it, and it can be read. maybe not easily, but if you don't wipe the drive, there is data on it.

But you can't just pull off files and pictures off the drive. There are only fragments of files on the disk.

Ah I see. I'm quite new to this so thanks for clarifying things!

This raised another question for me though. It is commonly said that if you lose more than 2 drives in a RaidZ2 array you will lose your entire pool and the data along with it. But if data a single drive can be read as you said, why would you lose all the data in the pool?

 

[Stux]

Because the array can't re-assemble the fragments.

Okay, imagine a house... of bricks... now remove 30% of the bricks then demolish it.

You try getting the house back.

The bricks are still there.

If you wanted to, you could reassemble the house... less 30%. It might take a while. It might cost a lot.

But more specifically, if you have a RaidZ2 array, and you lose 2 whole disks, then if you have another error you will lose data that is affected by that error. If that is a whole disk, you will lose the whole array. If its only 1 block which affects one file, you will only lose that file, but with raidz2 you will NOT lose any data until you suffer MORE than 2 failures affecting a given block. The first two disk failures affect every block. After that any other failure will wipe out what ever files are affected by that block... and if that block happens to be a critical disk metadata block, it could wipe out the filesystem, rather than a file. Since for all intents, a filesystem is actually just a file too ;)

With RaidZ1 if you lose a whole disk, then any further errors will cause data loss, and another total disk error will cause total data loss.

You should have a backup anyway

 

[nanoduckie]

Because the array can't re-assemble the fragments.

Okay, imagine a house... of bricks... now remove 30% of the bricks then demolish it.

You get the house back.

The bricks are still there.

If you wanted to, you could reassemble the house... less 30%. It might take a while. It might cost a lot.

But more specically, if you have a RaidZ2 array, and you lose 2 whole disks, then if you have another error you will lose data that is affected by that error. If that is a whole disk, you will lose the whole array. If its only 1 block which affects one file, you will only lose that file, but with raidz2 you will NOT lose any data until you suffer MORE than 2 failures affecting a given block. The first two disk failures affect every block. After that any other failure will wipe out what ever files are affected by that block... and if that block happens to be a critical disk metadata block, it could wipe out the filesystem, rather than a file. Since for all intents, a filesystem is actually just a file too ;)

Understood. Thanks for that clear explanation!

 

[Evi Vanoost]

Technically there is data on there that can be read. How much depends on your types of data and whether or not that is worrisome depends on your level of compliance you need to meet.

E.g. If you have a database of SSN technically it is possible to have many sets of SSN to fit in a single block on a drive. If it's a set of compressed videos, there may be small portions of it that may be recoverable, whether that is relevant depends on many variables. You may also expose full sets of metadata that describes more about the data your array contains.

If this is HIPAA-level data, then you should zero it out or trash the drives unless you want to pay for a statistician to prove it doesn't matter.

 

[Stux]

Also, if the blocks are compressed they might appear as encrypted anyway.