Re: [libvirt-users] live migration to localhost with lvm based storage

On 10/09/2014 03:01 PM, Vasiliy Tolstov wrote: So to restate and make sure I understand what you are asking, you have a running guest that you want to leave running, where the guest is currently using /path/to/vg1/lv1 (remote), and where you want to alter it to use /path/to/vg2/lv2 (local). If I'm correct, libvirt live block migration can do that; however, it requires that your domain be temporarily marked transient (due to the fact that the operation cannot be efficiently restarted; we hope to lift that restriction when some future qemu gives us better control over persistent dirty bitmaps). virsh dumpxml $dom > dom.xml virsh undefine $dom virsh blockcopy $dom /path/to/vg1/lv1 /path/to/vg2/lv2 --verbose --pivot edit dom.xml to change /path/to/vg1/lv1 to /path/to/vg2/lv2 virsh define dom.xml which will use qemu to copy the entire disk contents into a new location (that third step is potentially slow). Note that with new enough libvirt and qemu, you can get even fancier and faster, by using external commands that can move an lv between vgs with minimal copying time (I don't personally know lvm commands well enough to say what the sequence is, only that I suspect it can be done without the overhead of copying the entire lv by instead adjusting bookkeeping of which vg owns the lv): virsh snapshot-create-as $dom --no-metadata --disk-only \ --diskspec vda,file=/path/to/wrapper1.qcow2 work your magic to move/copy vg1/lv1 into vg2/lv2 qemu-img create -f qcow2 \ -o backing_file=/path/to/vg2/lv2,backing_fmt=qcow2 \ /path/to/wrapper2.qcow2 virsh dumpxml $dom > dom.xml virsh undefine $dom virsh blockcopy $dom vda --reuse-external \ /path/to/wrapper2.qcow2 --shallow --verbose --pivot edit dom.xml to change wrapper1 to wrapper2 virsh define dom.xml virsh blockcommit $dom vda --pivot --shallow --verbose -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org