> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rafftechnologies.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Database engines

> PostgreSQL, MySQL, Valkey, ClickHouse and Kafka on Raff, what each is for, versions, and what is different about each.

<sub>Updated October 11, 2026</sub>

Raff runs five engines. They share the same dashboard, API, plans and billing; what differs is what each is good at and a few engine-specific details.

## How it works

### PostgreSQL

The default choice for apps, APIs and AI. Versions 18, 17, 16 and 15; 18 when you do not choose.

* Two addresses: a **pooler** (PgBouncer, transaction mode) for app traffic and a **direct** connection for migrations and session features. See [Connection pooling](/products/store/databases/concepts/connection-pooling).
* Extensions such as `vector` (pgvector), `pg_trgm` and `pg_stat_statements`. See [Turn on extensions](/products/store/databases/quickstart-guides/turn-on-extensions).
* Point-in-time restore to any second, high availability with a standby, and read replicas.
* Database `defaultdb`, admin user `raffadmin`.

### MySQL

For WordPress, Laravel and apps already on MySQL. Version 8.0 (Percona Server).

* One address on port 3306.
* High availability runs a 3-node cluster (Galera) where every node holds all data.
* Point-in-time restore from the binary log.
* Database `defaultdb`, admin user `raffadmin`.

### Valkey

The open-source, Redis-compatible in-memory store, version 8. Any Redis client works over TLS (`rediss://`).

* Choose at create what happens when memory is full: **Cache** removes the least recently used keys; **Queue or data** keeps every key and refuses new writes.
* Snapshots every night; a restore can lose up to a day of keys.
* High availability adds a standby with automatic failover (Sentinel).

### ClickHouse

Column-oriented analytics: logs, events, metrics, dashboards over billions of rows. Version 25.8.

* Two addresses: HTTPS (8443) for HTTP clients and drivers, native TCP (9440) for `clickhouse-client`.
* High availability runs two replicas and a coordination service (ClickHouse Keeper). Use `Replicated*` table engines so tables are copied to both.
* Load data from CSV or Object Storage in the **Tables** tab.

### Kafka

Event streaming between services, version 4.3 (KRaft, no ZooKeeper).

* Private only: reach it from your VPC. Clients connect with SASL\_SSL and SCRAM-SHA-512 and must trust the database's CA certificate.
* High availability runs 3 brokers and copies every topic to all three.
* Topics are created on first use, and messages are kept 7 days by default.
* No backups: copies on the brokers protect the data.

## Which engine

| Need | Engine |
| - | - |
| A general app database, or vectors for AI | PostgreSQL |
| An app built for MySQL | MySQL |
| Cache, sessions, rate limits, job queues | Valkey |
| Analytics over large event or log tables | ClickHouse |
| Streams of events between services | Kafka |

## Trade-offs

* The engine and its major version are fixed after create. To move, create a new database and copy the data.
* The free plan covers PostgreSQL, MySQL and Valkey. ClickHouse and Kafka start at their smallest paid plan.

## Related

<CardGroup cols={2}>
  <Card title="Plans and sizing" icon="lightbulb" href="/products/store/databases/concepts/plans-and-sizing">
    Plans, connections and storage per engine.
  </Card>

  <Card title="Create a database" icon="rocket" href="/products/store/databases/quickstart-guides/create-a-database">
    Pick an engine and create it.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.