RADIUS Docker Image
This is yet another Docker FreeRADIUS 2 image with SQL support.
This image implements a sensible SQL schema with added tables to support period-based accounting.
By default, the SQL module for FreeRADIUS records the accounting start and accounting stop packets into a single row. This is a thoroughly convenient storage format, and very efficient for space. But it presents some challenges when you want to work out how much data has been used on a more granular period set; such as for a given hour within a single session.
Also, given sessions are much longer lived than they used to be (fibre based access services and even DSL have session times of weeks, months or years), if you need to do diagnosis on a session it can be hard to know whether the data consumed was this week, this month, or... any period of time.
This image extracts the accounting data every time an update is made and stores it so that you can process the data used in a period.
The SQL schema included implements an SQL trigger that extracts data updates every time an update to the accounting table is updated and stores it. There is then another CRON based script that summarises the data. The schema also includes some modifications to the standard FreeRADIUS schema based on the ARA FreeRADIUS management interface. This schema is compatible with ARA.
The image expects a MySQL server to be running and accessible from the hostname "mysql".
On boot this image attempts to connect to the MySQL server, and will loop until it can to load the SQL schema (if it's not there). If there is a database called 'radius' and it can connect using the default username and password (which are radius/radius) it will assume that the database exists and it does not need to be recreated. The init script requires the MySQL root password to be present in an environment variable (MYSQL_ROOT_PASSWORD).
The configurations aren't designed to be secure. This image is for development and testing purposes, and without some significant attention to security probably should not be used in production.