Hello, Marty
The volume format is determined by the hypervisor + storage type used.
Generally speaking, when working with KVM, file-based storages (like
NFS, shared mount point and local storage) use QCOW2, and other storage
types, such as RBD (Ceph,) use RAW. Could you share which storage type
you are using for your primary storage?
Furthermore, there are differences in implementation when using volume
and instance snapshots, could you share which type of snapshots are you
trying to take? I'm not familiar with snapshot creation when using RAW
volumes, but when using QCOW2 and file-based storage, the VMs should not
be paused during the whole process, at least in the latest versions.
Best regards,
João Jandre.
On 11/5/25 18:41, Marty Godsey wrote:
Hello,
On my KVM cluster, which is new and not in production yet since I am still
tuning, I noticed that snapshots were taking a long time to complete. And when
say a long time, I mean 45seconds to 1 min. The VM is, of course, paused when
this happens. As you can imagine, this is unusable. The connection to the NFS
storage is 100G and all NVME. While a snapshot is running, I can see it hitting
the storage really hard. I then looked at the KVM process for that VM on the
KVM server and it looks like it is using RAW for the format.
I created this VM by spinning up a new VM in CloudStack so I could make my
template.
So I have a couple questions. How do I specify QCOW2 when making new volumes or
spinning up new VMs and I assume that snapshots should not take this long,
correct?
Thank you for all your help.
