Thanks for letting me know I'm not making myself clear, let me try again with a few more specifics, These are Windows VMs with three disk images for C:, D: and L:. To simplify I'll focus on the image for C: which had grown to 77G physical size on the hypervisor's ssd (I just double-checked that with 'ls -al' because qemu-img below says 70G, this image had two snapshots at one time which may be the reason for the discrepancy). qemu-img info reports:
file format: qcow2virtual size: 100G (107374182400 bytes)disk size: 70Gcluster_size: 65536Format specific information:compat: 0.10
I used Windows Server 2012r2 "Optimize" (defrag) and then reduced the C: partition to about 67G in Disk Administrator leaving the remaining 33G as unallocated. Afterward I tried a web reference technique and used Sysinternals SDelete to zero the free space then used 'qemu-img convert -O qcow2 <original.qcow2> <new.qcow2>' to produce a physical image size of 29G. qemu-img info reports on "new.qcow2":
file format: qcow2virtual size: 100G (107374182400 bytes)disk size: 29Gcluster_size: 65536Format specific information:compat: 1.1lazy refcounts: falseThe issue is that the virtual size is still 100G. I don't have the physical disk space to allow approximately 12 images (all configured for a virtual size of 100G) to grow to that size (6 VMs with C: and D: on a 1TB ssd, L: is on an hdd which isn't an issue). I need to change that virtual size number for this image to 67 or 68G. On the D: images I can drop to approximately 40G for an aggregate total of about 650G for all six VMs - well within the 875G physical size limitation that the ssd provides after overhead.
That's a long background to why I need to change the virtual disk size. Again, any alternatives would be much appreciated.
-----Original Message-----
From: Martin Kletzander <mkletzan@redhat.com>
To: Leroy Tennison <leroy.tennison@verizon.net>
Cc: libvirt-users <libvirt-users@redhat.com>
Sent: Tue, Jul 11, 2017 2:51 am
Subject: Re: [libvirt-users] Is there still no easier way to shrink a VM image?
On Tue, Jul 11, 2017 at 12:34:31AM -0500, Leroy Tennison wrote:
>I have numerous qcow2 images which need to be reduced in size and have
>their maximum size (virtual size) reduced. Physical disk space became
>so low that VMs "auto-paused" themselves, I moved enough images to solve
>the immediate problem but need to rectify the underlying issue. It
>seems that qcow[2] files are grown in size such that the data inside of
>them takes about 50-60% of the space (does anyone know the actual
>algorithm or how to control it?). Given the total physical disk space
>on the hypervisors, I need something more restrictive.
>
I don't get it. You have virtual size greater than the free space on
the physical storage and instead of the VM finding out you want the
guest OS to see it has no space at all?
>Our hypervisors are a mix of Ubuntu 14 or 16 LTS (qemu-img 2.2 or 2.5).
>After doing all the preparation (defragment, reduce OS partition size)
>"qemu-img resize" reports that shrinking isn't supported yet. My web
>research indicates that, to accomplish this, I have to:
>
> convert to raw
>
> shrink the image
>
> convert back to qcow[2]
>
> increase the image size to provide for some growth
>
>I'm hoping I've missed something in my research and that someone knows
>an easier way. I don't feel constrained to qemu-img but this is a
>production environment precluding consideration of experimental
>software. Virt-resize, guestfish or any other reasonable option is fine
>with me. Solutions or ideas? Thanks.
>
>_______________________________________________
>libvirt-users mailing list
>libvirt-users@redhat.com
>https://www.redhat.com/mailman/listinfo/libvirt-users
_______________________________________________
libvirt-users mailing list
libvirt-users@redhat.com
https://www.redhat.com/mailman/listinfo/libvirt-users