On Sat, Oct 19, 2013 at 09:38:41AM +0300, Samuli Heinonen wrote:
Dear all,
i'm testing QEMU-Gluster features with oVirt 3.3.1 beta and RHEL 6.5 beta. For
reason or another it seems to bypass Gluster support in QEMU and use VM image through FUSE
mount point instead.
ie. Libvirt XML shown in vdsm.log looks like this:
<disk device="disk" snapshot="no" type="file">
<address bus="0x00" domain="0x0000"
function="0x0" slot="0x06" type="pci"/>
<source
file="/rhev/data-center/mnt/glusterSD/boar1.storage:dev-el6-sata1/402a2eee-40f6-4012-b188-63ec122ce399/images/f6ba1575-e6de-428c-93ef-8a27cf23de61/bc3de195-ccc4-4202-ade2-7e4d0cdb292e"/>
<target bus="virtio" dev="vda"/>
<serial>f6ba1575-e6de-428c-93ef-8a27cf23de61</serial>
<boot order="1"/>
<driver cache="none" error_policy="stop"
io="threads" name="qemu" type="raw"/>
</disk>
and also qemu-kvm process for this VM has option "-drive
file=/rhev/data-center/mnt/glusterSD/boar1.storage:dev-el6-sata1/402a2eee-40f6-4012-b188-63ec122ce399/images/f6ba1575-e6de-428c-93ef-8a27cf23de61/bc3de195-ccc4-4202-ade2-7e4d0cdb292e,if=none,id=drive-virtio-disk0,format=raw,serial=f6ba1575-e6de-428c-93ef-8a27cf23de61,cache=none,werror=stop,rerror=stop,aio=threads"
oVirt 3.3 and Fedora 19 works as intended. Any ideas what might be wrong?
Steps to reproduce (hopefully?):
1: Clean install of ovirt-engine 3.3.1 beta on CentOS 6.4
2: Clean install of RHEL 6.5 beta with EPEL, Gluster-epel and oVirt repos enabled (and
all necessary Red Hat channels.)
3: Install RHEL 6.5 host to oVirt and let it install necessary packages
4: Switch datacenter storage type from NFS -> GlusterFS
5: Insert storage domain
6: Start installing virtual machines
Please let me know if more information is needed or you want me to open a bug report for
this.
As we've learned during today's ovirt weekly meeting, this is probably a
manifestation of
Bug 1022961 - Running a VM from a gluster domain uses mount instead
of gluster URI