ZFS: create dataset in pool or not?

janvv

Active Member
Jun 21, 2020
64
12
28
67
52.24154182722349, 5.117853866801705
I am new to ZFS, but because I have set up my server with a couple of equal-sized disks, I think it makes sense to create RAID1 sets with them. Obviously, I can do that with LVM, but I want to learn something new.

I created a ZFS pool on the command line:

Bash:
zpool create nvme_pool mirror /dev/sdb /dev/sdc

This creates a directory /nvme_pool as a mount point. I could add that directory as storage in the Proxmox datacenter. Ready. Simple.

But... to learn more about ZFS, I watched some videos on YouTube. Almost everyone is doing an extra step: creating an additional mount point, like /nvme_pool/nvme or whatever name they use. That mount point is then used to define a new storage.

I can't get my head around why that would be beneficial. The only reason I can think of—but I haven't seen that in the videos—is to create separate datasets for ISOs, backups, disk images, etc.

Can experienced ZFS users shed some light on my ZFS confusion?


UPDATE:
In the meantime, I am convinced to create datasets in the pool.
For others having similar doubts, go and read this Reddit thread:
https://www.reddit.com/r/zfs/comments/wvu3g4/are_datasets_necessary_for_a_single_disk/
 
Last edited: