ZFS pool question

A

approx

New Member
Jan 18, 2026
4
0
1
I have never used ZFS and want to understand if my plan is possible.

I have 1 4TB NTFS external hard drive (proxmox is installed on an internal drive, this 4TB is just for data storage nothing system related) ATM. My plan was to get a 8TB externalhard drive, format it and make a ZFS pool with it and then transfer all the data from the 4TB drive onto the 8TB ZFS drive. Then I’d format the 4TB drive and add to the ZFS pool, making it 12GB total.

Is this possible with ZFS? If so, can I continue to add drives in this manner indefinitely?
 
Possible, but that would make it a RAID0 (a single stripe in ZFS terms) and you might lose all data when one of the drives fails. This can also be done with LVM and BTRFS.
Anyway, be careful as you double your chances to lose data from a drive failure.
 
  • Like
Reactions: Kingneutron and approx
leesteken said:
Possible, but that would make it a RAID0 (a single stripe in ZFS terms) and you might lose all data when one of the drives fails. This can also be done with LVM and BTRFS.
Anyway, be careful as you double your chances to lose data from a drive failure.
Click to expand...
Thanks. Is there a way to do it without that risk or is that inherent in the nature of what I’m trying to do?
 
approx said:
Is there a way to do it without that risk or is that inherent in the nature of what I’m trying to do?
Click to expand...
If you spread out data over N drives then you have roughly N-times as much chance of a drive failure. If you don't have redundancy (you want 8+4=12) then there is no duplicate copy of the data and it is gone when the only copy is lost (due to a drive failure). Lots of information about RAID-levels on the internet as this is not Proxmox specific.
 
  • Like
Reactions: Kingneutron
Using a ZFS mirror would provide redundancy but 4 TB space. Sometime later, you can replace the smaller drive with an 8+ TB and "grow" the pool to 8 TB.

The speed of the external drive could also be a concern because writes would write to both.

There are ways to achieve redundancy and drive independence. Ceph, in Proxmox, but that's going to require a minimum 3 PVE nodes to be useful, as well as high speed networking, and drives on each node. Windows has Storage Spaces which requires/uses multiple drives but obviously that doesn't help here.
 
  • Like
Reactions: approx
SteveITS said:
Using a ZFS mirror would provide redundancy but 4 TB space. Sometime later, you can replace the smaller drive with an 8+ TB and "grow" the pool to 8 TB.

The speed of the external drive could also be a concern because writes would write to both.

There are ways to achieve redundancy and drive independence. Ceph, in Proxmox, but that's going to require a minimum 3 PVE nodes to be useful, as well as high speed networking, and drives on each node. Windows has Storage Spaces which requires/uses multiple drives but obviously that doesn't help here.
Click to expand...
I’m now looking at alternatives like mergerfs. It has a fill up setting where it fills one disc and then moves on to the next but it still appears as a single volume in the OS. I need to do more research but it looks promising. Obviously that’d mean I’d still lose one disc worth of data in the event of a failure, but that’s better than losing everything. I do plan to expand storage one day to include back up but I’m not in a position to atm
 
That's very similar to LVM, which is actually supported. If you don't need the other features of ZFS or BTRFS, just use LVM where only one file or virtual disk will be partly on one drive and party on the other (or use two separate VG to prevent this), and the rest will be on either drive. Don't over-complicate it, is my suggestion and make sure you have good backups (I prefer PBS) anyway.
 
  • Like
Reactions: Kingneutron and approx
leesteken said:
That's very similar to LVM, which is actually supported. If you don't need the other features of ZFS or BTRFS, just use LVM where only one file or virtual disk will be partly on one drive and party on the other (or use two separate VG to prevent this), and the rest will be on either drive. Don't over-complicate it, is my suggestion and make sure you have good backups (I prefer PBS) anyway.
Click to expand...
Thanks, I just read about LVM linear. That sounds perfect :) I appreciate your help!
 
I need to chime in...
approx said:
Thanks, I just read about LVM linear. That sounds perfect
Click to expand...
You did not talk about your use case, only "just for data storage". I would accept "perfect" only if the data has absolutely zero value. (Or you run daily automatic backups, at least.)

If the data is valuable then you may want to have some features I listed there: https://forum.proxmox.com/threads/f...y-a-few-disks-should-i-use-zfs-at-all.160037/

Again: your mileage may vary and LVM has advantages and may fit. But you should know what makes ZFS w/ redundancy so interesting...
 
Last edited:
  • Like
Reactions: news
Individual external USB drives are relatively unreliable. They have no active cooling, can get bumped or knocked over, and you have external power and data cables that need to remain solidly connected. Then there is the higher chance of finding an SMR drive in an external unit, rather than CMR. All of that adds up to a higher chance of data loss.

As a thought experiment involving only a 4TB and 8TB drive, I would probably use the 8TB drive as primary storage and then the 4TB drive for daily or weekly backups of the most critical data via script or some manual method.

I have used MergerFS a fair amount with my OMV+MergerFS+SnapRAID server and I consider all three of those to be tough solutions to recommend except for specific use cases. Specifically for MergerFS, I think you'd find it a bit maddening to select a write policy that makes sense for a 4TB+8TB drive arrangement. You can write data to the combined share, but you'll not necessarily know where the files are going unless you manually check or one drive fails.

Overall, I don't think you're providing enough information about what you're trying to accomplish. So these suggestions are going to be relatively inaccurate or more general in nature.
 
You must log in or register to reply here.
Top Bottom
Back