octavia

Bug #1558368
Comment #2

Comment 2 for bug 1558368

Revision history for this message

Leslie Lundquist (llundquist-e) wrote on 2016-06-06:

Getting Started with Installing Octavia Q & A

Q. Hello, I have a question regarding using Octavia. I have a running setup with controller node, network, and hypervisors. We got LBAASv2 working with HAproxy, installing LBaaSv2 on the controller and lbaas-agent and haproxy on the network host.

For Octavia, where exactly is the module installed? On controller node, network, or on whichever hypervisor we want to deploy our load balancing VM?

Also, do we need a VM image for Octavia to deploy, and if so, how do we prep it? What needs to be preconfigured on the image?

A. Octavia boots service VMs to do the load balancing. The Octavia controller processes reside on your OpenStack controller nodes. Octavia does not require to be installed on specific nodes.

As for the service VM image, we provide a script that will build an image in /diskimage-create under the Octavia directory. Our test gates use "diskimage-create.sh -s 2" to create a 2GB image.

Q. So we do not need to install any additional stuff on our network hosts or hypervisors (as we did with HAproxy)?

A. Right, yes, the HAproxy drive has to be on the network node.
You do need to install Octavia somewhere in your OpenStack.

To use Octavia, instead of installing the LBaaSv2 agent and HAproxy, you should install Octavia control processes (API, worker, health manager, and housekeeping)

Does Octavia deploy the load-balancing VM automatically when we create a load-balancer/listener (similar to the way HAproxy and agent creates an qlbaas Linux namespace)?
A. Yes, Octavia handles booting the VM.

Q. 1 more question :) How does it know to use the image created via the above script? We have many images in our image library. Is the image name/path specified somewhere in the LBaaS CLI or API?

A. You provide the image id or tag in the octavia.conf file.

Getting Started with Installing Octavia Q & A

For Octavia, where exactly is the module installed? On controller node, network, or on whichever hypervisor we want to deploy our load balancing VM?

Also, do we need a VM image for Octavia to deploy, and if so, how do we prep it? What needs to be preconfigured on the image?

A. Octavia boots service VMs to do the load balancing.  The Octavia controller processes reside on your OpenStack controller nodes. Octavia does not require to be installed on specific nodes.

As for the service VM image, we provide a script that will build an image in /diskimage-create under the Octavia directory. Our test gates use "diskimage-create.sh -s 2" to create a 2GB image.

Q. So we do not need to install any additional stuff on our network hosts or hypervisors (as we did with HAproxy)?

A. 	Right, yes, the HAproxy drive has to be on the network node.
	You do need to install Octavia somewhere in your OpenStack.

To use Octavia, instead of installing the LBaaSv2 agent and HAproxy, you should install Octavia control processes (API, worker, health manager, and housekeeping)

You can install them anywhere you want, as long as Neutron can reach the Octavia API.
Q. Yes, I was just confused as to whether we needed a separate Octavia module running somewhere on our hosts in addition to a controller node or a node where neutron-server lies.
	
Does Octavia deploy the load-balancing VM automatically when we create a load-balancer/listener (similar to the way HAproxy and agent creates an qlbaas Linux namespace)?
A. Yes, Octavia handles booting the VM.

Q. 1 more question :) How does it know to use the image created via the above script? We have many images in our image library. Is the image name/path specified somewhere in the LBaaS CLI or API?

A. You provide the image id or tag in the octavia.conf file.