On 12/07/2012 01:22 PM, Michael Morgan wrote:
Hello,
I've been building an OpenStack cluster using Ceph as the storage backend.
Currently running SL 6.3, OpenStack Folsom packages from EPEL, Libvirt 0.9.10,
QEMU 1.2.1 built from an epel-testing SRPM for RBD support, and Ceph 0.54. I
can boot off of an image without problems but attaching to a running instance
fails. Tracking back from Horizon and Cinder I see Libvirt generating the
errors:
2012-12-07 20:03:20.657+0000: 19803: error : qemuMonitorJSONCheckError:338 : internal error unable to execute QEMU command 'device_add': Property 'virtio-blk-pci.drive' can't find value 'drive-virtio-disk1'
2012-12-07 20:03:20.664+0000: 19803: error : qemuMonitorTextDriveDel:2895 : operation failed: deleting file=rbd:volumes/volume-6c82e5d3-e697-43a8-8194-1f7df932ceb8:id=volumes:key=AQCicbZQ8Oo2IBAALganQv+zY/jjECc9fHUBBA==:auth_supported=cephx none,if=none,id=drive-virtio-disk1,format=raw,cache=none drive failed: drive_del: extraneous characters at the end of line
I can reproduce the error with virsh attach-device and the following XML:
<disk type='network' device='disk'>
<driver name='qemu' type='raw' cache='none'/>
<auth username='volumes'>
<secret type='ceph' uuid='93fb3d32-7e2d-691d-db94-4c1cf21bed02'/>
</auth>
<source protocol="rbd" name="volumes/volume-6c82e5d3-e697-43a8-8194-1f7df932ceb8"/>
<target bus='virtio' dev='vdb'/>
</disk>
I realize this is probably more of a Libvirt/QEMU problem but the only
reference I could find to an error like this was a post here from last year:
http://permalink.gmane.org/gmane.comp.file-systems.ceph.devel/4713
In that thread the problem disappeared after multiple version changes without
any clear source for the resolution. Has anyone seen this type of error before?
Yes, this is fixed in libvirt 0.9.12. There were a couple issues with
properly escaping JSON and using a space instead of a semicolon with
qemu's command parsing.
There are a couple more rbd-related bug fixes in later versions of
libvirt, but you wouldn't hit those problems using OpenStack.
I can certainly move along to other lists if appropriate but figured I'd start
here since it's the only place I saw such an issue pop up. It also gives me an
opportunity to say how awesome Ceph seems. The level of support I've seen on
this list is pretty amazing and I hope to increase our use of Ceph in the
future.
Thanks :). I actually prefer ceph-related stuff coming to this list,
since it's easier to keep track of and gets more attention from people
more familiar with ceph, rather than just libvirt, qemu, etc.
Josh
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
