Setting up your database

Mailman uses the SQLAlchemy ORM to provide persistence of data in a relational database. By default, Mailman uses Python’s built-in SQLite3 database, however, SQLAlchemy is compatible with PostgreSQL and MySQL, among possibly others.

Currently, Mailman is known to work with the SQLite3, PostgreSQL, and MySQL databases. (Volunteers to port it to other databases are welcome!). If you want to use SQLite3, you generally don’t need to change anything, but if you want Mailman to use PostgreSQL or MySQL, you’ll need to set those up first, and then change a configuration variable in your /etc/mailman.cfg file.

Two configuration variables control which database Mailman uses. The first names the class implementing the database interface. The second names the URL for connecting to the database. Both variables live in the [database] section of the configuration file.

SQLite3

As mentioned, if you want to use SQLite3 in the default configuration, you generally don’t need to change anything. However, if you want to change where the SQLite3 database is stored, you can change the url variable in the [database] section. By default, the database is stored in the data directory in the mailman.db file. Here’s how to tell Mailman to store its database in /var/lib/mailman/sqlite.db file:

[database]
url: sqlite:////var/lib/mailman/sqlite.db

PostgreSQL

First, you need to configure PostgreSQL itself. This Ubuntu article may help. Let’s say you create the mailman database in PostgreSQL via:

$ sudo -u postgres createdb -O $USER mailman

You would also need the Python driver psycopg2 for PostgreSQL:

$ pip install psycopg2

You would then need to set both the class and url variables in mailman.cfg like so:

[database]
class: mailman.database.postgresql.PostgreSQLDatabase
url: postgres://myuser:mypassword@mypghost/mailman

If you have any problems, you may need to delete the database and re-create it:

$ sudo -u postgres dropdb mailman
$ sudo -u postgres createdb -O myuser mailman

Many thanks to Stephen A. Goss for his contribution of PostgreSQL support.

MySQL

First you need to configure MySQL itself. Lets say you create the mailman database in MySQL via:

mysql> CREATE DATABASE mailman;

You would also need the Python driver pymysql for MySQL.:

$ pip install pymysql

You would then need to set both the class and url variables in mailman.cfg like so:

[database]
class: mailman.database.mysql.MySQLDatabase
url: mysql+pymysql://myuser:mypassword@mymysqlhost/mailman?charset=utf8&use_unicode=1

The last part of the url specifies the charset that client expects from the server and to use Unicode via the flag use_unicode. You can find more about these options on the SQLAlchemy’s MySQL page.

If you have any problems, you may need to delete the database and re-create it:

mysql> DROP DATABASE mailman;
mysql> CREATE DATABASE mailman;

Database Migrations

Mailman uses Alembic to manage database migrations. Let’s say you change something in the models, what steps are needed to reflect that change in the database schema? You need to create and enter a virtual environment, install Mailman into that, and then run the alembic command. For example:

$ python3 -m venv /tmp/mm3
$ source /tmp/mm3/bin/activate
$ python setup.py develop
$ mailman info
$ alembic -c src/mailman/config/alembic.cfg revision --autogenerate -m
  "<migration_name>"
$ deactivate

This would create a new migration which would be applied to the database automatically on the next run of Mailman.

People upgrading Mailman from previous versions need not do anything manually, as soon as a new migration is added in the sources, it will be automatically reflected in the schema on first-run post-update.

Note: When auto-generating migrations using Alembic, be sure to check the created migration before adding it to the version control. You will have to manually change some of the special data types defined in mailman.database.types. For example, mailman.database.types.Enum() needs to be changed to sa.Integer(), as the Enum type stores just the integer in the database. A more complex migration would be needed for UUID depending upon the database layer to be used.