OpenStack Compute (nova)

regading msg https://lists.launchpad.net/openstack/msg12296.html

Asked by Vidya

I checked the ssues which Sebastain has. and even I got device busy error.
 It looks nova-volume creates an xml which is sent to libvirt without the hostname and ip.

Connection.py logs in nova
2012-06-21 16:25:38 INFO nova.virt.libvirt.connection [req-1801ba4a-02ce-4486-98eb-68248a8ef320 9a52e76e98514d9099e96e9f7db4b133 d09adc7cdc0e41ba9f9a5172cc7a0997] <disk type='network'>
                         <driver name='qemu' type='raw' cache='none'/>
                         <source protocol='rbd' name='nova/volume-00000001'/>
                         <auth username='admin'>
                             <secret type='ceph' uuid='8d07be7e-ab3e-5181-b900-adc261be7c44'/>
                         </auth>
                         <target dev='vdd' bus='virtio'/>
                     </disk>

Once i chnage volume.py and driver.py to include ceph hostname. It seems to work
<driver name="qemu" type="raw" cache='none' />
  <source protocol="rbd" name="nova/volume-00000004">
    <host name='192.168.7.10' port='6789'/>

  </source>
  <target dev="vdb" bus="virtio"/>
  <auth username='admin'>
    <secret type='ceph' uuid='8d07be7e-ab3e-5181-b900-adc261be7c44'/>
  </auth>
</disk>

I was not able to pots my message in that issue

Is it a valid issue that nova-volume not creating a proper xml or any config issue on nova/ceph side?

Question information

Language:
English Edit question
Status:
Expired
For:
OpenStack Compute (nova) Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Launchpad Janitor (janitor) said : 		#1

This question was expired because it remained in the 'Open' state without activity for the last 15 days.

Revision history for this message
Josh Durgin (jdurgin) said : 		#2

Right now the RBD driver in OpenStack requires you to have your monitor addresses in /etc/ceph/ceph.conf. When you have them there, putting them in the libvirt xml is unnecessary. You do have to make sure that the unix user who ends up running qemu/kvm has read access to /etc/ceph/ceph.conf, and e.g. apparmor isn't preventing qemu/kvm from reading it.

This dependency on /etc/ceph/ceph.conf should be removed in the future.

Revision history for this message
Josh Durgin (jdurgin) said : 		#3

Update for those finding this in the archives: In Havana, cinder sends the monitor addresses to nova, which includes them in the instance xml, so having /etc/ceph/ceph.conf on a compute node is no longer necessary.