Description of problem: QEMU now provides the efficient way to create images (raw,qcow2) on glusterfs volumes directly using libgfapi and not using fuse mount. (e.g) qemu-img create gluster://gluster-server/vol-name/file1.img 30G qemu-img create gluster://gluster-server/vol-name/dir1/file1.img 30G But there are no options currently available to make use of such an image available in glusterfs volume(using this qemu+glusterfs driver ), for the creation of VM This bug is raised to add such an option Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: No option provided for making use of qemu+glusterfs driver, for the creation of VM Expected results: There should be an option to make use of qemu+glusterfs driver Additional info: qemu-system-x86_64 has option called "--file", which expects the argument in the form of, "gluster://<gluster-server-host>/<gluster-volume-name>/[<dirname>/]filename" Note: directory name is optional
virt-manager does not want to ever run any qemu tools directly. This requires adding support for a gluster based storage pool in libvirt first.
Well there's domain <disk> support for gluster volumes, right? I figured that's what the initial IRC query was about. But yes if you want virt-install to create a new volume and not just use an existing one, it will require libvirt patches as well.
Also, we should at least make sure virt-manager UI doesn't blow up or look awful when it sees a network disk
*** Bug 531420 has been marked as a duplicate of this bug. ***
Another bit, from the duped bug, is allowing HTTP cdrom installs. I don't think we should add an explicit new mode for it, but maybe we can make virt-install/virt-manager 'just work' if passed an HTTP iso path as CDROM media. Something to explicitly test at least.
Upstream virt-install supports: --disk $URL for gluster, http, etc. It should 'just work'. It also allows explicit building like: --disk source_protocol=gluster,source_name=...,...