[SOLVED] Allocating a virtual disk on a zpool

[EllyMae]

I have an almost empty zpool onto which I want to allocate a 15TB zvol. There is an existing 15TB zvol on another storage that I want to move to the new zpool to relieve some IO contention.

The existing zvol looks like this after a zfs list -p command.
NAME USED AVAIL REFER
fivebays/vm-105-disk-1 15052666255184 19656901499904 15052666255184

The destination zfs filesystem looks like this:
NAME USED AVAIL REFER
threeamigos 36634726469 19803765370811 130944

It appears that there is enough space on the destination. But, when I, say, ask Proxmox to move the disk from fivebays to threeamigos, I get an out-of-space message.

create full clone of drive scsi1 (Zvol:vm-105-disk-1)  [Zvol is the name of the Proxmox storage that points to fivebays.]
TASK ERROR: storage migration failed: zfs error: cannot create 'threeamigos/vm-105-disk-1': out of space

Why am I getting this message?

 

[leesteken]

You are probably using raidz1 (or 2 or 3) and this has a lot of overhead and padding with typical volblocksizes. Several threads about that on this forum but this is also a good overview: https://web.archive.org/web/2021031...or-how-i-learned-stop-worrying-and-love-raidz

 

[EllyMae]

I would think that the overhead and padding between the two almost-identically-configured zpools (except for the number of drives) would be the same. Therefore, the space used in the first zpool should be the same as the space used in the second zpool.

But that's all in the past, albeit unsolved, since I've moved on to an alternate way to achieve my overall goal.

 

[EllyMae]

I think the correct number for the capacity of a ZFS dataset is given by the zfs list command. Am I right?

 

[LnxBil]

I think the correct number for the capacity of a ZFS dataset is given by the zfs list command. Am I right?

There are no absolutes. It depends most importantly on the padding overhead that does indeed correspond to the number of drives in a vdev and the used volblocksize.