RAID Array with different sized disks on a NAS server

[DeanNotDin]

Hey!

 

I finally managed to sell my Western Digital EX4, the most awful home NAS server I ever had the joy to own.

 

Now I managed to get my hands on a spare PC, and installed FreeNAS 9.3.

 

My disk setup is currently 2x4TB (WD Red), 1x2TB (WD Purple) and 1x512GB (WD Blue).

The pool i'm looking to set up would be used mainly for home media storage and streaming, but also for daily backups for the computers at home.

 

What would you recommend as the best ZFS pool setup, with balance between performance and redundancy (1 Disk if enough, I believe.)

[Oshino Shinobu]

Your only option, if you're wanting some redundancy, is to RAID 1 the two 4TB Red drives. You'd end up wasting drive capacity using the other drives in the array and it's not advised to mix drives in an array anyway. 

 

Personally, I would recommend having all of the drive separate and then backing up all of the important data on an external drive. 

[Electronics Wizardy]

You could use something like btrfs or storage spaces or a pool with parity to make a volume with different sized disks. 

[DeanNotDin]

On 2/7/2017 at 0:21 AM, Oshino Shinobu said:

Your only option, if you're wanting some redundancy, is to RAID 1 the two 4TB Red drives. You'd end up wasting drive capacity using the other drives in the array and it's not advised to mix drives in an array anyway. 

It does mean just wasting a lot of storage for nothing, And I do know that its not advised. None the less, doing so is indeed an option.

 

On 2/7/2017 at 0:21 AM, Oshino Shinobu said:

Personally, I would recommend having all of the drive separate and then backing up all of the important data on an external drive. 

I do like that approach, and it might be what I will end up doing.

But I have to say that the entire point of having a NAS server is avoiding manually backing up data to a hard drive.

On 2/7/2017 at 0:39 AM, Electronics Wizardy said:

You could use something like btrfs or storage spaces or a pool with parity to make a volume with different sized disks. 

This is exactly my question, how will I be able to set up a pool with parity on different sized disks.

And for what I know, ZFS is more stable than btrfs, and I wouldn't want to use storage spaces.

[Electronics Wizardy]

28 minutes ago, DeanNotDin said:

And for what I know, ZFS is more stable than btrfs, and I wouldn't want to use storage spaces.

Basically what you do is use something line union fs to make a group of disks look like one volume. Then use snap raid to make a parity on the largest drive and run the parity sync. This is basically what unraid does and let's you combine the capacity of all the drives. 

 

Btrfs is stable in raid 1 0 and 10 and will let you easily add a disk when you want and supports checksums. 

[Fpmurphy]

Your best option with ZFS is to either make a pool with the 2x4TB drives in a mirrored vdev or combine your disks in a Raid Z1.

 

If you go with the mirrored vdev route, you can always add other combinations of drives as vdevs to expand. However, you typically want to stick with the same type. So if you put the 2x4TB drives in a mirror, you'll want to add other drive pairs as mirrors. You could add the 2TB and the 512GB as a second mirrored vdev to your pool. This would give you a total of ~4.5TB in your pool.

 

ZFS doesn't handle different sized disks very well. It will limit any vdev to the size of the smallest drive in your pool. If you make a Raid Z1 vdev with all of your disks, it will be treated as a 4X512GB vdev. Until you replace the 512GB disk with a third 4TB drive. Then it will be treated as a 4x2TB vdev.

 

If you're looking to just be able to add different sized drives as you get them, look into unraid or another btrfs based OS.

[DeanNotDin]

On 2/8/2017 at 1:28 AM, Electronics Wizardy said:

Then use snap raid to make a parity on the largest drive and run the parity sync. This is basically what unraid does and let's you combine the capacity of all the drives. 

Doesn't that mean that the implementation is kind of the same as RAID 4, in which only a single drive serves as a parity drive?

In that case this drive will "die" sooner because of the repetitive parity writes, and also this will cost me the space of the biggest drive in the array?

[Electronics Wizardy]

21 minutes ago, DeanNotDin said:

Doesn't that mean that the implementation is kind of the same as RAID 4, in which only a single drive serves as a parity drive?

In that case this drive will "die" sooner because of the repetitive parity writes, and also this will cost me the space of the biggest drive in the array?

Yep its basically file level raid 4.

 

 

I don't think drive life spawn will be a huge issue. 

 

You will have to use the biggest drive in the array for this, and can't normally use it for any other data.

[dalekphalm]

9 hours ago, DeanNotDin said:

Doesn't that mean that the implementation is kind of the same as RAID 4, in which only a single drive serves as a parity drive?

In that case this drive will "die" sooner because of the repetitive parity writes, and also this will cost me the space of the biggest drive in the array?

You're starting to see why using any sort of RAID-like system for the drives you have isn't really a good system. At best, you'd get 6.5TB of actual raw storage (Less, once converted from TB into usable TiB).

 

As for the risk of the parity drive dying? Not a super large risk. Certainly, there will be more wear and tear on the drive, but if it's a good drive, it'll still last you years. I ran FlexRAID for years, and I used a system very much like this. I did, however, match my drive sizes (virtually, in any case), as follows:

 

1x 3TB HDD (dedicated parity drive)

 

1x 3TB HDD

 

1x virtual 3TB HDD (made up of)

2x 500GB

1x 2TB

 

I eventually added another 3TB HDD to the mix, but anyway, the dedicated parity drive is still in use today in a different build. It was a Seagate 3TB ST3000DM001.