47d82ffac7
Includes diagram of what a Libra based LBaaS solution should look like Change-Id: Ia5caec0a6b0d566a1a348dc90f9478df2fbfb4db
30 lines
1.0 KiB
ReStructuredText
30 lines
1.0 KiB
ReStructuredText
Introduction
|
|
============
|
|
|
|
Libra is a Load Balancer as a Service (LBaaS) system originally designed by
|
|
Hewlett-Packard Cloud Services. It consists of three of the core components
|
|
required to get LBaaS working:
|
|
|
|
* A command line client
|
|
* A node pool manager to keep a warm spare pool of load balancers ready
|
|
* A node worker to asyncronusly communicate to the API server
|
|
|
|
It does not (yet) include the API server itself or HAProxy. The API server is
|
|
based on the Atlas API system but communicates to the workers using gearman.
|
|
|
|
Layout
|
|
------
|
|
|
|
.. image:: libralayout.png
|
|
|
|
Here you can see that the pool manager spins up the required Nova nodes with
|
|
the load balancer image. It then hands the details of this node over the the
|
|
API server.
|
|
|
|
The client sends an Atlas API request to the API server which in-turn sends the
|
|
configuration information to the worker on the load balancer node. The worker
|
|
has a plugin API to speak to multiple load balancer types but is currently
|
|
designed to use HAProxy.
|
|
|
|
The parts of this diagram in orange are provided by the libra codebase.
|