Hard Disk Alocation Issues with thin provisioning based disks (.VMDK and .QCOW2)

[fecosco]

The problem: The hard disk clone or import function allocate 100% nominal disk size.
PVE Version: 6.2-10
Background: .VMDK and .QCOW2 disks are thin provisioning based. This means that if you create a new 20gb disk, the disk allocation will grow until it reach 20GB.
The problem: Wile clone or import .VMDK or .QCOW2 disks the PVE convert the disks allocate 100% nominal size even the real internal disk allocation is lower than its value. For instance: I have a Linux machine with 128gb hard disk and I only installed the CENTOS 7 plus some tools on it. The real alocation is about 8GB but while cloning this disk the new disk usage is 100% or 128GB. This is killing my storage management.
The paliative solution: I use the "qemu-img convert" command, but is exhaustive and time consuming do it every time I clone or import a new disk to the PVE.



Question: There any workaround that could be done to solve this behavior?

 

[ertanerbek]

Make tamplate and use linked clone... Also discard and discard after cloning option good for you.

 

[fecosco]

Make tamplate and use linked clone... Also discard and discard after cloning option good for you.

Hi @@ertanerbek, thanks for advice. I'm already working with templates with no issues. The problem is that every time I import new disk or when I need to full clone some VM I need to perform those manual steps to fix the allocation issues. I think that something must be done by proxmox developers to solve this issue, If the thin technology disk has 7.8gb of data sounds crazy when it allocate all 128gb nominal space.

Regards!

 

[ertanerbek]

if you activate discard feature on your guest, when you start your guest then guest disk allocation space will be down but that means also more IO