On 01/21/2013 04:35 PM, Nux! wrote:

On 21.01.2013 13:41, Gary Kotton wrote:

...

On 01/21/2013 03:27 PM, Nux! wrote:

...

Hi,

I want to clarify, on a controller/db only server, what quantum services do I need to run? Or do all services need to run on al hypervisors and only use the DB of the controller?

Please take a look at http://docs.openstack.org/trunk/openstack-network/admin/content/connectivity...:

On the controller you only need to run the Quantum server. If the plugin has an agent then the agent will need to be run on the compute or network nodes.

Thanks Gary _______________________________________________ cloud mailing list cloud@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/cloud

Thanks Gary, that image is quite clear regarding the services, but it also raises another question, I see one can run the quantum agents on a dedicated machine. What would be the benefit of that versus running quantum-agents on all hypervisors? (assuming everybody wants to do SDN/OVS/gre). It's a bit confusing.

True, it is very confusing.

The quantum-*-agents (if they exist - some plugins may not have agents - for example controllers may talk directly with the OVS switches) do the following: they attach a interface to the network. This may be assigning tags etc. There are two cases when this is done. The first is when the hypervisor creates a VM, for example libvirt will create a tap device and attach the device to the bridge. The quantum agent will ensure that this tap device has the correct VLAN tag. This takes place on the compute node. The second case is when the DHCP agent or the L34 agent create tap devices - for example allocating IP addresses. These devices need to be attached to the quantum network. Hence the quantum-*-agent will need to run on the network node.

Hope that this helps.

Thanks Gary