Open edX Platform Database and Search Technologies

Open edX Platform Database and Search Technologies#

The Open edX platform uses several databases and search backends, each serving different purposes.

MySQL: The long-standing primary relational database for most services (LMS, CMS/Studio), storing users, courses, enrollments, and grades.
PostgreSQL: Newly supported as of the Verawood release, with migrations updated to make the Open edX platform compatible across both MySQL and PostgreSQL. Changes include dynamic SQL generation based on the database engine, cross-database field type handling, and custom psycopg2 adapter registration for Open edX’s OpaqueKey types like CourseLocator. A community-developed Tutor plugin (tutor-contrib-postgresql) supports PostgreSQL-based Tutor deployments. See Use PostgreSQL as the Relational Database Backend.
SQLite: Used in local development and testing environments.

MongoDB: Used to store course content structure in Studio. There is ongoing work to migrate away from MongoDB toward MySQL via the Learning Core (openedx-learning) initiative.

Redis: Used for caching, session storage, and Celery task queuing. Superseded Memcached in modern deployments.
Memcached: Legacy caching layer, largely replaced by Redis.

Meilisearch: The current primary search engine for content library and course authoring search (introduced in Quince), used by the Authoring MFE for course and library content discovery, as well as the Catalog MFE for course searching.
Typesense: A supported alternative search backend for site operators, attractive in part because it supports raft-based clustering for high availability — a limitation that has been noted with Meilisearch. See Use Typesense search backend.

Note

Typesense is currently (as of Verawood) only supported for LMS searching, _not_ Studio/libraries.
Elasticsearch / OpenSearch: Used for older LMS-level search (course catalog, user search). OpenSearch is the current preferred variant, with ongoing community discussion about whether to deprecate it in favor of lighter-weight alternatives.

Redis (also serves this role) or RabbitMQ: Used as a Celery broker for background task processing.

Review Date	Working Group Reviewer	Release	Test situation
2025-04-10	sarina	Verawood	Pass