Elasticsearch Was Never a Database

4 months ago

Elasticsearch was designed as a search engine, not a primary database.
Using Elasticsearch as a primary database leads to issues with transactions, schema migrations, and query limitations.
Elasticsearch lacks atomic transactions, making it unreliable for operations requiring consistency.
Schema changes in Elasticsearch often require reindexing, which can be risky when it's the primary store.
Elasticsearch's query capabilities are limited compared to relational databases, especially for joins.
Durability and recovery in Elasticsearch are not as robust as in traditional databases.
Operating Elasticsearch at scale introduces operational challenges and instability.
Misusing Elasticsearch as a database increases complexity, cost, and engineering effort.
Elasticsearch should remain a search engine, with the primary database handling the system of record.
ParadeDB is presented as an alternative combining OLTP and full-text search in one system.

Hasty Briefsbeta