[SOLVED] How much data does a RAID 5/6 rebuild need to read?

[Pyneappel]

How much data needs to be read in order to rebuild a RAID array after a single drive failure? Does it have to read all the data on all the other disks even on larger (say 16 or 24 drive arrays)? I'm specifically interested in ZFS RAIDZ2 of at least 8 drives but will take info for generic raid6 or raid5. Or, does it only have to read the equivalent of, say, three other disks, be that three physical disks or blocks equivalent to three disks' worth of data spread throughout the array?

 

[popatim]

None. Raid 5 & 6 both remain operational if a single drive fails. The rebuild happens in the background, you can set the rebuild performance level in your raid controllers interface, but users will typically notice a performance decrease.

 

[Pyneappel]

I understand the array is still accessible with one disk failure. that isn't my question. My question is; in order to re-create the content of the failed disk, the system needs to read and XOR data off of other disks (correct?) ... the only examples I can find are the most simplistic ones of, say, what happens in a 3-disk RAID 5. What I don't understand is in an array with a larger number of disks how many are used in the parity calculation and how the data is striped or sharded across the disks. In a 10-disk raid 5 does the system use the data on all 9 of the other disks to create the parity bit for the 10th? Etc.

 

[USAFRet]

During the rebuild, it reads and works with ALL the data and ALL the drives.

How long it takes depends on how much data, among other things.
When I rebuilt the 4 drive RAID 5 in my QNAP, moving from 4x 3TB to 4x 4TB....swapping in a single new drive took about an hour per TB for the rebuild.