Database Schema Management¶
This page describes how database schema updates are managed by the Inmanta core.
Definition new schema version¶
A new version of the database schema is defined by adding a new Python module to the
inmanta.db.versions package. The
name of this module should have the format
v<version>.py, where <version> is an integer indicating the version of
the new database schema. Version numbers start at 1.
Each of these Python modules should implement an asynchronous function
update that accepts a database connection object
as an argument. This function should execute all database queries required to update from the previous version of the
database schema (<version> - 1) to the new version of the database schema (<version>). All changes done by the update
function should be executed in the transaction. An example is given in the code snippet below.
# File: src/inmanta/db/versions/v1.py async def update(connection: Connection) -> None: schema = """ ALTER TABLE public.test ADD COLUMN new_column; """ async with connection.transaction(): await connection.execute(schema)
Executing schema updates¶
Schema updates are applied automatically when the Inmanta server starts. The following algorithm is used to apply schema updates:
Retrieve the current version of the database schema from the
public.schemamanagertable of the database.
Check if the
inmanta.db.versionspackage contains any schema updates.
When schema updates are available, each
updatefunction between the current version and the latest version is executed in the right order.
When a schema update fails, the database schema is rolled-back to the latest schema version for which the
did succeed. In that case the Inmanta server will fail to start.