I don't know of any technical limitations within QEMU (although most likely untested), but it definitely appears that libvirt restricts you to a file or block device. Note that we do have a librbd backlog item for built-in image migration [1] that would follow a similar approach as QEMU.
On Fri, Oct 20, 2017 at 10:24 AM, Li Wang <laurence.liwang@xxxxxxxxx> wrote:
Hi，
We want to use the drive mirror provided by
qemu to mirror a local qemu image to a ceph
pool for backup, and vm migration. The usage
likes 'virsh blockcopy instance_id vda /cephfs/vm.qcow2', We have tested CephFS, it woks. And we have done a simple try for kernel rbd,
namely, use '/dev/rbd0' instead of '/cephfs/vm.qcow2', it seems works too (right?).
However, we think the most efficient way
with the shortest io path is to use qemu-rbd
driver, to not reach host kernel at all. We are
wondering if it is feasible, or need some
development?
Cheers,
Li Wang
发自网易邮箱大师
Jason
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
