Our current goal is to use only two means of communication between Wazo processes:
- a REST API over HTTP for synchronous commands
- a software bus (RabbitMQ) for asynchronous events
Each component should have its own REST API and its own events and can communicate with every other component from across a network only via those means.
The current xivo-dao Git repository contains the basis of the future services Python API. The API is split between different resources available in Wazo, such as users, groups, schedules... For each resource, there are different modules :
- service: the public module, providing possible actions. It contains only business logic and no technical logic. There must be no file name, no SQL queries and no URLs in this module.
- dao: the private Data Access Object. It knows where to get data and how to update it, such as SQL queries, file names, URLs, but has no business logic.
- model: the public class used to represent the resource. It must be self-contained and have almost no methods, except for computed fields based on other fields in the same object.
- notifier: private, it knows to whom and in which format events must be sent.
- validator: private, it checks input parameters from the service module.
The goal is to make Wazo as elastic as possible, i.e. the Wazo services need to be able to run on separate machines and still talk to each other.
To be in accordance with our goal, a Wazo daemon must (if applicable):
- Offer a REST API (with encryption, authentication and accepting cross-site requests)
- Be able to read and send events on a software bus
- Be able to run inside a container, such as Docker, and be separated from the Wazo server
- Offer a configuration file in YAML format.
- Access the Wazo database through the
- Have a configurable level of logging
- Have its own log file
- Be extendable through the use of plugins
- Not run with system privileges
- Be installable from source
- Service discovery with consul
Currently, none of the Wazo daemons meet these expectations; it is a work in progress.