FreeNAS Without the RAID

[Philbar715]

I have several differently sized disks from 3TB to 8TB. I would like to add them all to a single "pool" so that they appear as a single large drive. I do not need the availability of RAID, and I do not need a backup of the data that is stored on this volume. Is this possible with Freenas? Similar to Stablebit Drivepool (but without the mirror).

 

[danb35]

Is this possible with Freenas?

Of course it is, you'd just stripe them together into your pool. Be aware that when one of those disks fails, all your data on the entire pool would be lost.

 

[Jailer]

I have several differently sized disks from 3TB to 8TB. I would like to add them all to a single "pool" so that they appear as a single large drive.

You can do this but they will all appear the size of the smallest disk. In your case they would appear to be a group of 3TB drives.

 

[kdragon75]

@danb35 OS correct. @Jailer is correct for drives added to a single vdev. In your case, each drive would be its own vdev so this is not an issue. Just remember that the pool reliability will be that of one drive devided by the number of disks.

 

[Allan Jude]

You can do this but they will all appear the size of the smallest disk. In your case they would appear to be a group of 3TB drives.

If you just stripe disks together, they you will get all of the storage space from all of the disks. However, the risk is extremely high, and this is not a recommended configuration.

One possible configuration is to create mirror pairs, of different sized disks, if you have 4 disks: 3,4,5,8
3+4 = 3TB mirror
5+8 = 5TB mirror
= 8TB usable space

The advantage is that in the future, if you replace the 5TB with another 8TB, the usable space of that mirror will go from 5 to 8TB.

You could then use that 5 in place of the 3, and the 2nd mirror would grow from 3 to 4TB (as it is now a 4TB and a 5TB disk)

 

[Jailer]

Yeah I misunderstood the OP's question. Setting the disks up as individual vdev's is what he's looking for but as everyone has suggested that's a bad idea.

 

[Philbar715]

thanks for the help guys! Im only considering this option because I dont care if i lose the data thats on these disks (all my important files and docs have a separate backup system). I would prefer not to have to stripe the disks together but thats what I'll probably end up doing.

 

[kdragon75]

thanks for the help guys! Im only considering this option because I don't care if i lose the data thats on these disks (all my important files and docs have a separate backup system). I would prefer not to have to stripe the disks together but thats what I'll probably end up doing.

The only other way is disk concatenation and that's such a horrible crappy way of doing things almost nothing supports it including ZFS. What's wrong with striping them?

 

[Philbar715]

The only other way is disk concatenation and that's such a horrible crappy way of doing things almost nothing supports it including ZFS. What's wrong with striping them?

I would prefer to have easy expand-ability down the line by just adding more disks, and I would prefer that if I lost one drive I only lost the data that was on that disk instead of the entire array.

 

[garm]

I’m confused, either you care about the content or you don’t?
If you don’t care about the content pool all the disks individually. If you don’t want to loose all the data on a disk failure, make each disk it’s own pool. But unlike “normal” file systems ZFS will tell you when a file is broken and you will not be able fix it unless you have some level of redundancy. It’s probably better not to know in that case. You should proabobly look for a different file system

 

[kdragon75]

I would prefer to have easy expand-ability down the line by just adding more disks, and I would prefer that if I lost one drive I only lost the data that was on that disk instead of the entire array.

I don't get it. You happy to play russian rulet with random files but not all of them? As Garm mentioned, ZFS is all about knowing when something is wrong and being able to fix it. With a big stripe, you still get the checks just no repair. But you also get all the performance of all of the disks. With concatenation, you in theory could get the check but no way to fix an error and there is no way of knowing what you would lose is a disk goes. Either have sain expectations for your mix bad of hardware or get proper hardware. Perhaps if we knew more about your collection of disks we could make a compromise suggestion? ie you have 1 3tb 4 6tb and 1 8tb, great toss the 3tb and use a raidz1. As you need more space, add disks 6tb or bigger. Run out of ports? replace all 6tb drives with 8tb drives and gain the difference. Otherwise I don't think you want to use ZFS.

 

[RueGorE]

... and I would prefer that if I lost one drive I only lost the data that was on that disk instead of the entire array.

... I would like to add them all to a single "pool" so that they appear as a single large drive...

It doesn't seem like you're aware what you're asking for, because if you added all your disks into a single pool, the entire pool would be lost if any disk were to fail. If you wanted data to be lost only on its respective disk (vdev), then you'd create separate pools dedicated to each vdev.

 

[pro lamer]

I would prefer not to have to stripe the disks together but thats what I'll probably end up doing.

Maybe some old FreeNAS version allows this (one supporting UFS) but since I am a noob neither I know if it is possible nor how dangerous an advice to use an old system is...

And moreover I don't know if it is what kdragon75 meant "crappy" :)

The only other way is disk concatenation and that's such a horrible crappy way of doing things almost nothing supports

Sent from my phone

 

[danb35]

No version of FreeNAS, at least back to the 0.6 days, allows anything like this. unRAID, I think, does.

 

[sretalla]

NAS4Free (seems now to be calling itself xigmanas) does UFS, so FreeNAS had it at least for some time before the fork.

Although NAS4Free is in my opinion far inferior to FreeeNAS it's an option if you want to depart from the ethos of ZFS, but still want the option to use it at some point.

 

[danb35]

Yes, old FreeNAS (actually through the 9.2 series) did UFS, but it didn't support the kind of disk spanning that OP seems to be talking about. And yes, NAS4Free has now renamed itself to XigmaNAS because reasons.

 

[pro lamer]

Is that true that you want FreeNAS because of ZFS checksums? Or because of some other features?

Some dumb idea: make many single disk pools, make some main parent directory somewhere and one directory per disk, and symlink them to the parent directory ;P

Sent from my phone

 

[garm]

Or nullfs the crap out of it, as Matt Damon might have said

 

[Arwen]

Linux supports something called UnionFS that combines individual file systems, (which could be EXT3/4 or ZFS single disk pools), into what appears to be a single file system. I have no experience with UnionFS, so I could be wrong. And it does not appear to be included with the normal kernel sources.

Edit: Looks like UnionFS is available for FreeBSD, (but I don't know if it's available in FreeNAS). There are even some alternatives listed;

Wikipedia - UnionFS

 

[kdragon75]

Is that true that you want FreeNAS because of ZFS checksums? Or because of some other features?

Some dumb idea: make many single disk pools, make some main parent directory somewhere and one directory per disk, and symlink them to the parent directory ;P

Sent from my phone

The limitation there is that a single file would not be able to span two drives along with the space per "folder" limits that would impose. It just comes back to the fact that this is not an appropriate way of building storage systems and that is made clear by the fact that Microsoft is the only one the readly supports it. We all know who good Microsoft is at storage...