MySQL

MySQL (8.0+) is a first-class detectkit backend. Like ClickHouse it uses databases (MySQL has no separate schema concept). MySQL 8.0 is required for the row-alias upsert detectkit uses to deduplicate internal state.

Install

pip install "detectkit[mysql]"   # driver: pymysql

profiles.yml

default_profile: dev

profiles:
  dev:
    type: mysql
    host: localhost
    port: 3306
    user: root
    password: ""
    internal_database: detectkit   # database for detectkit's own _dtk_* tables
    data_database: analytics       # database your metric source tables live in

  prod:
    type: mysql
    host: "{{ env_var('MYSQL_HOST') }}"
    port: 3306
    user: "{{ env_var('MYSQL_USER') }}"
    password: "{{ env_var('MYSQL_PASSWORD') }}"
    internal_database: detectkit
    data_database: monitoring

Field	Required	Notes
host / port	yes	default port 3306
user / password	yes	the user needs CREATE to auto-create the databases
internal_database	yes	detectkit auto-creates it (CREATE DATABASE IF NOT EXISTS)
data_database	yes	database your source tables live in
database	no	optional default database for the connection
settings	no	extra pymysql.connect keyword arguments

Requires MySQL 8.0+. detectkit deduplicates internal state with the row alias form INSERT ... AS new ON DUPLICATE KEY UPDATE, introduced in 8.0.19.

MariaDB

MariaDB is fully supported through this same MySQL backend — same driver (pymysql), same connection fields, same table/DDL layout. The one internal difference: the row-alias upsert form above (INSERT ... AS new ON DUPLICATE KEY UPDATE) isn’t available on MariaDB, so detectkit auto-detects the server vendor at connect (SELECT VERSION()) and falls back to the classic VALUES(col) form there instead:

INSERT INTO t (...) VALUES (...)
ON DUPLICATE KEY UPDATE col = VALUES(col), ...

You don’t configure this — it’s picked automatically per connection.

Use the type: mariadb alias for clarity (identical semantics to type: mysql):

profiles:
  prod:
    type: mariadb
    host: "{{ env_var('MARIADB_HOST') }}"
    port: 3306
    user: "{{ env_var('MARIADB_USER') }}"
    password: "{{ env_var('MARIADB_PASSWORD') }}"
    internal_database: detectkit
    data_database: analytics

type: mysql against a MariaDB server works too — vendor detection is by the live server, not the profile type.

pip install "detectkit[mariadb]"   # same pymysql driver as detectkit[mysql]

dtk init --db-type mariadb scaffolds a project with a type: mariadb profile. Tested against MariaDB 11.x in the integration matrix; MariaDB 10.4+ is expected to work.

Metric query dialect

Use MySQL SQL. The equivalent of a bucketed aggregate:

SELECT
  FROM_UNIXTIME(FLOOR(UNIX_TIMESTAMP(event_time) / {{ interval_seconds }})
                * {{ interval_seconds }}) AS timestamp,
  SUM(status_code >= 500) AS value
FROM http_requests
WHERE event_time >= '{{ dtk_start_time }}' AND event_time < '{{ dtk_end_time }}'
GROUP BY 1
ORDER BY 1

How detectkit stores state

Internal tables have an enforced primary key; detectkit deduplicates with a version-aware INSERT ... ON DUPLICATE KEY UPDATE (newest row wins), reproducing the ClickHouse ReplacingMergeTree guarantee. JSON columns are stored as TEXT; use JSON_EXTRACT(...) if you query them in dashboards.

See the Databases overview and Profiles.