Probably asked before, i can't find a link or topic.

 

Anyway, i can't get my head round the idea that without using any striping or spreading of data over all disks in the total array, it is possible to physically remove one disk, and then let the system rebuild that disk from the relatively small parity disk like unRAID claimes.

 

If i have an unRAID array of 20 disks of 3TB each, from which ONE is a 3TB parity disk, and i remove one of 19 data disk, how is it technically possible to rebuild the missing 3TB from the parity that is collected over the entire 60TB??? The parity itself will max out at 3TB, and that is for the WHOLE array.

 

With traditional RAID the data is basically spread over the total array, so in previous example,  if i would place a 20KB file on the volume, each disk would get 1KB. So if one disk fails, i'm just missing 1KB in stead of the total 20KB and have still 19KB left + parity... But with unraid, i will miss the entire 20KB, only the small parity data is left. I can't imaging that the small parity that is left, is enough for re-creating the entire disk...

 

Or is the parity spread over the total array? Or does each disk has parity of the other disks but its own?

How does it work? What am i missing here?

 

*edit* didn't look far enough... i think i got it the (even!) parity is calculated over matching blocks over ALL disks at once... so with 20 disks, if 1 disk is missing, you just need to reverse the parity calulation over the other 19 disk block by block to re-create the missing one

 

Pretty simple once you 'see' it

Yep. So simple it confuses people. 

 

Now that you understand it, you should also realize how important it is to only use drives that are tested and trusted completely. When one drive fails, all the rest are called on to read back perfectly. Never use a questionable drive "because it's protected from failure". You'll end up with 2 failed drives before you know it.