Kolla design philosophy¶
This page aims to explain the design choices that have been made in the Kolla projects (the main Kolla project for building the images as well as all subprojects aimed at deploying them).
It is written down mostly for potential contributors but may be a good guideline for anyone using and integrating with Kolla projects as it gives an overview of what the preferred approach is. As noted, this is of great interest to contributors because it might explain why certain changes will get rejected and why some others will need a remake in order to be accepted.
This page is likely an always-in-progress, living document. Please reach out to the team via the mailing list or the IRC channel to discuss the rules noted here, see our docs on communication for details.
Do not own the deployed services’ config defaults (unless necessary)¶
In Kolla, we try not to own service config and its defaults. We believe the upstream services know best what works for them by default.
This rule is overridden when a certain config option needs to be orchestrated among multiple services or otherwise needs to agree with certain other assumptions made in the environment. The notable exceptions here are the basic addressing of services and Keystone credentials.
Prefer documented configuration via config overrides¶
… as opposed to new in-Ansible variables.
This plays nicely with the config overrides capabilities of Kolla Ansible which let users easily customise the services’ config to their hearts’ contents regardless of what Kolla Ansible offers via Ansible variables.
The reasoning behind this rule is that it lowers the maintenance burden yet it does not handicap the users - they can control every last detail of their config.
One of the (not necessarily planned) side-effects of this is that users’ config overrides are not locked-in for Kolla Ansible and can relatively easily be reused in other configuration systems.