[SOLVED] Confused about discrepancy in reported disk usage for same drive

kik0

New Member
Feb 29, 2024
2
0
1
Trying to understand two different usage reports for the same drive in Proxmox 8.4.11

When I look at Host > Disks > ZFS it says I have 832GB free on my 1TB nvme drive (Picture1)
When I look at the Storage section for the same drive, it shows I have 488GB free (Picture2)

Not sure I understand the discrepancy of about 344GB between the two reports.
 
Last edited:
I have the same question, did you figure out the reason?
From what I can tell, under Host > Disks > ZFS, the values represent your physical drive’s total capacity (Size) and the portion that is still available (Free) after subtracting what is already reserved (Allocated).

For example, on my system I have a 1 TB NVMe used for VMs and LXCs. It shows 179 GB allocated (though it’s unclear to me what exactly is holding that space) and 817 GB free, which seems to be the usable space available for new VMs or LXCs.

When you look under Storage for the same drive, that view reflects the actual usage by the system (VMs, LXCs, etc.). In my case, it shows I’m currently using 521 GB of the 1 TB drive.

What remains a bit of a mystery is what the 179 GB “allocated” represents and which process or system component is reserving it.
 
With ZFS, everything is more complicated due to its dynamic recordsize, thin-provisioning, compression and deduplication features. Usage or free space cannot be computed so easily, especially the padding overhead. This always leads to confusion.

There are two commands that operate at different levels: zpool and zfs, so the difference in the UI come down to the ouput of those two views on your storage.

There are a lot of threads on the forums that deal with this exact problem, e.g.
 
Thanks guys! This post is what explained my discrepancy: https://forum.proxmox.com/threads/proxmox-storage-different-than-zfs-usage.142095/. I hadn't enabled Thin Provision in Datacenter > Storage initially, so the VMs had USEDREFRESERV set on their volumes still.

Doing zfs set refreservation=none pool/volume made the graph in Host > Storage show the correct disk usage total for the datasets, and new VM disks have USEDREFRESERV 0 now as well, due to enabling the Thin Provision checkbox, I believe.
 
Last edited: