Thank you!

I created all the VMs using the sparce allocation method.  I wanted a method that would create disks that did not immediately occupy their full declared size (eg, allow overcommit of disk space, as most VM hard drives are 30-50% empty for their entire life).

I kinda figured that it would not free space on the underlying storage when a file is deleted within the disk.  What confuses me is a disk that is only 30GB to the OS is using 53GB of space on gluster.  In my understanding, the actual on-disk usage should be limited to 30GB max if I don't take snapshots.  (I do like having the ability to take snapshots, and I do use them from time to time, but I usually don't keep the snapshot for an extended time...long enough to verify whatever operation I did was successful).

I did find the "sparcify" command within ovirt and ran that; it reclaimed some space (the above example of the 30GB disk which is actually using 20GB inside the VM but was using 53GB on gluster shrunk to 50GB on gluster...But there's still at least 20GB unaccounted for there.

I would love it if there was something I could do to reclaim the space inside the disk that isn't in use too (eg, get that disk down to just the 21GB that the actual VM is using).  If I change to virtio-scsi (its currently just "virtio"), will that enable the DISCARD support, and is Gluster a supported underlying storage?

Thanks!
--Jim

On Fri, Sep 1, 2017 at 5:45 AM, Yaniv Kaul <ykaul@redhat.com> wrote:


On Fri, Sep 1, 2017 at 8:41 AM, Jim Kusznir <jim@palousetech.com> wrote:
Hi all:

I have several VMs, all thin provisioned, on my small storage (self-hosted gluster / hyperconverged cluster).  I'm now noticing that some of my VMs (espicially my only Windows VM) are using even MORE disk space than the blank it was allocated.

Example: windows VM: virtual size created at creation: 30GB (thin provisioned).  Actual disk space in use: 19GB.  According to the storage -> Disks tab, its currently using 39GB.  How do I get that down?

I have two other VMs that are somewhat heavy DB load (Zabbix and Unifi); both of those are also larger than their created max size despite disk in machine not being fully utilized.

None of these have snapshots.

How come you have qcow2 and not raw-sparse, if you are not using snapshots? is it a VM from a template?

Generally, this is how thin provisioning works. The underlying qcow2 doesn't know when you delete a file from within the guest - as file deletion is merely marking entries in the file system tables as free, not really doing any deletion IO.
You could run virt-sparsify on the disks to sparsify them, which will, if the underlying storage supports it, reclaim storage space.
You could use IDE or virtio-SCSI and enable DISCARD support, which will, if the underlying storage supports it, reclaim storage space.

Those are not exclusive, btw.
Y.


How do I fix this?

Thanks!
--Jim

_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users