5tb

oh ok so that's why i currently have 8.13tb volume on my 4x3tb NAS

I don't understand how it is possible to backup 8TB of data on only 4TB of drive
I suppose if two drives fails then i am totally screwed ?

Even if only 4TB is used to "backup" 8TB of data, do i still get the same data protection than if i kept a full duplicate of my datas (50% disk space for files and 50% for backup) ?

anarchoi wrote:
oh ok so that's why i currently have 8.13tb volume on my 4x3tb NAS I don't understand how it is possible to backup 8TB of data on only 4TB of drive

Not backup, redundancy. A backup is stored on a separate device preferably at a different location.

This is not the best illustration but you could think of it like this:

Let us call the four disks A, B, C and D

Disk A = 1/3 B + 1/3 C + 1/3 D
Disk B = 1/3 A + 1/3 C + 1/3 D
Disk C = 1/3 A + 1/3 B + 1/3 D
Disk D = 1/3 A + 1/3 B + 1/3 C

If any one disk fails your data remains intact as between your other disks you have a copy of all the data on the failed disk.

anarchoi wrote:
I suppose if two drives fails then i am totally screwed ?

You would need to try cloning one of the disks but if the disks are too far gone then you would need to restore data from backup.

anarchoi wrote:
Even if only 4TB is used to "backup" 8TB of data, do i still get the same data protection than if i kept a full duplicate of my datas (50% disk space for files and 50% for backup) ?

RAID-10 still has the issue that two disks failures can result in data loss.

RAID is great and RAID (except RAID-0) helps minimise the risk of data loss. There are problems that RAID can't protect against. That's why backups to another NAS or some place else are important.

mdgm's illustration doesn't quite work for me. Here's another way to think about the redundancy. Probably more than you want to know. :D

With single redundancy and 4 disks, the volume is divided into groups of blocks. There are 3 data blocks and one parity block in each group. Each of these blocks are on a different disk.

The parity block is computed using xor

A xor B xor C => P
You can reconstruct any missing block from the other three.

A = B xor C xor P
B = A xor C xor P
C = A xor B xor P
P = A xor B xor C

This works if (a) there is only one block missing and (b) you know which block that is. That lets you solve the proper equation. If more than 1 block is missing in the group, then you are screwed. If you don't know which block is bad, you are also screwed. When you replace a drive, the NAS reconstructs all the blocks that are now missing from the remaining drives using the appropriate equations above. That is what a resync does (and is why a resync fails if there are any bad blocks on the other drives).

RAID-6 uses two parity blocks, the second one is computed using a different formula. That creates simultaneous equations, so you can reconstruct any 2 missing blocks from the rest. Like Raid-5 you need to know which two are missing, and if more than 2 blocks are missing in the group you are screwed.

On the sparc ReadyNAS all the P blocks on the the last drive in the array. The newer NAS spread the P blocks across all the drives (using a different striping pattern). Every write requires rewriting the P block, so you get better performance if you spread them across all the drives.

BTW, one consequence of cloning is that the NAS thinks all the blocks on the cloned drive are correct. Since that is not really the case (some of the original blocks couldn't be read), it will result in some data corruption.

thanks to both of you for those explanations, i understand now :)

The only part i don't understand is:

BTW, one consequence of cloning is that the NAS thinks all the blocks on the cloned drive are correct. Since that is not really the case (some of the original blocks couldn't be read), it will result in some data corruption.

If i understand correctly, cloning is only used when 2 drives fails and it will result in some data corruption. If only 1 drive fail, there will be no data corruption ?

A long time ago i replaced 2 of the drives in my NAS at same time (i assumed it was the 2 backup drives because i was thinking the drives were 50% files 50% backups) and i noticed some movies were impossible to read after that. I guess that's why.

Cloning is only needed if you have multiple disk failures.

If two drives fail then the volume would fail to mount.

Cloning a drive e.g. using dd_rescue may help to get most of your data back. Though it would be best to seek advice from tech support before proceeding to clone a disk.

mdgm wrote:
Cloning is only needed if you have multiple disk failures...Cloning a drive e.g. using dd_rescue may help to get most of your data back...

Absolutely- getting back something is much better than losing everything. Definitely there are cases where cloning is your best option.

But you likely will lose something, and it will be hard to figure out exactly what that something is.

For me the main thing is not to depend on RAID for more than it can deliver. It's very convenient to be able to expand the volume and replace failed disks without losing access to the data. But it won't protect your data from loss as well as a backup will.