Howto setup quantum for floating IPs similar to the way nova-network did it?

Asked by Nick Bartos

I'm trying to get quantum floating IPs working in this configuration. Nova-network was working fine before, but now we are switching to quantum for better configurability in some installations. Here is a breif description of the setup:

The only router for all openstack networks is a (real) 10G ethernet switch. There is one untagged network used for internal communication (not exposed to the VMs). There are 2 additional networks (tagged VLANs on the same physical network), one for VM fixed addresses, and one for public floating addresses. All hosts have access to all networks, and no hosts have dedicated tasks (there are no networking-only, compute-only, etc hosts, but services that there can be only one/few of are master elected). Currently quantum is using the linux bridge plugin, but that will likely change to something else later.

The NAT for the floating IPs needs to be done on the host where the VM resides. Having a separate networking-only NAT host will not work in our configuration. This is because we do not want to bottleneck network I/O through any one host, and also so that if one host fails it will not affect the network on other surviving hosts. From what I can tell by the docs, a quantum router for NAT may need to be created for each host to make this happen, although I'm not really sure.

Currently quantum-server is a master elected service on a floating IP (which is managed outside of openstack), so it runs on only one node at a time. The dhcp, linuxbridge, and l3 agents are running on each host.

How exactly should quantum be set up to do this? Or is this configuration not feasable using quantum?

Question information

Language:
English Edit question
Status:
Answered
For:
neutron Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
dan wendlandt (danwent) said :
#1

Hi Nick,

Unfortunately, as we describe in the limitations section (http://docs.openstack.org/trunk/openstack-network/admin/content/ch_limitations.html), a deployment model similar to nova's --multi_host flag is not supported in quantum folsom.

There is work targeted for grizzly-3 that will support co-locating a router on the compute host along with a VM, see:
- https://blueprints.launchpad.net/quantum/+spec/quantum-scheduler
- https://docs.google.com/a/nicira.com/document/d/1TJlW0_tMpeENA_ia38fvRu7ioKRt9fsWXBjivwd1mMw/edit

Can you help with this problem?

Provide an answer of your own, or ask Nick Bartos for more information if necessary.

To post a message you must log in.