If no one objects, I'll clean the logging table in the next days.

@jcrespo Any safe way to just clean data older than one month?

I'm going to re-ping @jcrespo and @Marostegui because I'm unsure about how to do this properly and without danger. Issuing DELETE or DROP commands are dangerous. If you could give me ideas on how to do this I'd appreciate it. Thanks.

For what I see, this is a user database (meaning it is not a replicated mediawiki table, right?). If that is the case, there is very little risk. If you want to be sure you do not overload the server, using pt-archiver is a good way to do smaller transactions, even if it takes more time.

Some context:

MariaDB [s51541_sulwatcher]> describe logging;
+-------------+------------------+------+-----+----------------+----------------+
| Field       | Type             | Null | Key | Default        | Extra          |
+-------------+------------------+------+-----+----------------+----------------+
| l_id        | int(10) unsigned | NO   | PRI | NULL           | auto_increment |
| l_regex     | varchar(255)     | NO   | MUL | NULL           |                |
| l_user      | varchar(255)     | NO   |     | NULL           |                |
| l_project   | varchar(255)     | NO   | MUL | NULL           |                |
| l_timestamp | binary(14)       | NO   | MUL | 19700101000000 |                |
+-------------+------------------+------+-----+----------------+----------------+
5 rows in set (0.00 sec)

MariaDB [s51541_sulwatcher]> SELECT COUNT(*) FROM logging;
+----------+
| COUNT(*) |
+----------+
|   216107 |
+----------+
1 row in set (0.44 sec)

In addition to Jaime's comments, when issuing drops, it is generally a good idea to do them with "if exists":
drop table xx if exists
drop database xx if exists

This would avoid replication issues (if there was replication) with tables or databases under a master-slave model that might not be in sync for whatever reason.

Mentioned in SAL (#wikimedia-cloud) [2017-12-04T10:24:04Z] <TabbyCat> Fixed T151113 by cleaning data older than 20171201000000.