This is a libvirt question, so asking it on the qemu-block might not get you an answer that quick, ... or ever if I didn't notice your other question also addressed incorrectly. [adding libvirt-users@xxxxxxxxxx to cc] On Tue, Sep 01, 2020 at 10:57:34 -0400, Yoonho Park wrote: > I am trying to perform a blockpull on an attached disk, but I cannot do > this using "virsh attach-disk" specifying a new, empty overlay file created > with "qemu-img create". The VM does not pick up the backing_file path from Could you please elaborate on the exact steps? specifically the arguments for qemu-img create. My feeling is that you didn't specify the backing image format (-F) when creating the overlay, but it's hard to tell from this very sparse question. > the qcow2 overlay file, and there seems to be no way to specify the backing > path through the attach-disk command itself. Instead the backing path has virsh attach-disk is a syntax-sugar wrapper which generates the XML document which is in fact used to attach disk. You can use virsh attach-disk --print-xml to see what would be used. > to be specified in the xml file used to create the VM, which precludes > attaching a new disk to a running VM. Is this a bug? Is it possible to using virsh attach-device allows you to pass in a whole <devices> subelement such as <disk> with any configuration you are able to specify when defining the XML. The virsh attach-disk is limited to just "basic" config options. > attach a disk to a running VM, and specify its backing path, using qemu > directly? This is with qemu 4.2.0 and libvirt 6.0.0.
