Newbie General Questions

[mjk2374]

I'm looking to build out an Unraid Array. I have a few simple questions.

 

I know that the parity drive should be the largest drive. Does this place a cap on the drive size in the array once it is setup?

 

So if I build an array today with 3 - 3TB drives (one as parity and two as regular disks) and buy a 4TB at a later date, that 4TB must be setup as a parity, correct?

 

I'm a little confused as to how redundancy can happen with a large array that isn't striped with a single tiny parity disk. For instance if I had 10 - 3TB drives each with their own filesystem, how could I rebuild a failed drive with a 3TB parity disk?

 

I understand how it works in RAID, but I'm not sure about unRAID. It seems like data loss is a given with a failed disk. I really don't want an array that requires a restore for redundancy. Backups should be backups.

 

I would like to primarily use unRAID as an NFS fileserver with a minimal amount of smb shares. I'd like to see all of the disks as a combined mountable disk on the NFS clients. So if I had an 9 TB array, I get an 9 TB NFS volume to mount on the NFS clients. This as opposed to several smaller disks. Can I accomplish this task with NFS and unRAID?

 

Thanks in advance!

 

[garycase]

Yes, that's correct.

 

[JonathanM]

Unraid doesn't keep individual drive recovery information on the parity disk, it uses all the disks that haven't failed, plus the parity drive, to calculate the contents of a SINGLE failed drive. Let me state that another way. There is NO useful information on the parity drive, except when combined with ALL the other disks. There isn't a valid file system on the parity drive.

 

Read here, and here.

 

If you lose access to 2 drives, you lose the contents of both drives. Hopefully in that case one of the failed drives is the parity drive, since it has no usable info on it.

[garycase]

There are several tutorials on how parity works ... it's really very simple:

 

For any given bit, all of the disks are "added" in a one-bit binary register ...

 

e.g. Say Disk1 had a 1, Disk2 had a 0, Disk3 was a 1, Disk4 was a 1

 

Then you'd get 1+0+1+1 = 1  (Remember 1+0 =1, 1+1=0)

 

UnRAID uses "even parity", so the parity bit is always the value necessary to make the sum of all bits in the array a 0.  So in this case, the parity bit would be a 1.  It doesn't matter how many drives there are -- as long as no more than one drive has failed, you can restore its contents by simply adding up all the bits from the other drives and computing whether that drive had been a 0 or a 1.

 

That's why the parity drive limits how large the drives in the array can be .. if it wasn't you could only protect the data up to the size of the parity drive.

 

[mjk2374]

I think you just solved the mystery I had in my head with this one quote "There isn't a valid file system on the parity drive". 

 

 

Unraid doesn't keep individual drive recovery information on the parity disk, it uses all the disks that haven't failed, plus the parity drive, to calculate the contents of a SINGLE failed drive. Let me state that another way. There is NO useful information on the parity drive, except when combined with ALL the other disks. There isn't a valid file system on the parity drive.

 

Read here, and here.

 

If you lose access to 2 drives, you lose the contents of both drives. Hopefully in that case one of the failed drives is the parity drive, since it has no usable info on it.