Scaling

Hyperflask is made of different processes. It uses honcho as process manager.

Process	Description	Defaut port	Can be multiplied
web	The application server handling HTTP requests	5000	Yes
worker	The worker that executes background tasks		Yes
scheduler	The worker that will trigger cron jobs		No
mercurehub	The mercure hub server for SSE	5300	No

Each of these processes can be started independently.

• web processes can be served by a load balancer proxy (like nginx)
• worker processes can be hosted on different machines and multipled to handle more tasks

Important

You won't be able to use SQLite as a database once you reach scaling issues. We recommend switching to PostgreSQL.

Starting individual processes

Using Hyperflask-Start and the provided Dockerfile to build a container:

$ docker run my-app-image run web # only start the web process
$ docker run my-app-image run worker scheduler # only start the worker and scheduler processes

Scaling workers

If you are hosting workers on another machine than the app server, a queue server will be needed.

Dramatiq, the background tasks library at the heart of Hyperflask, can use different brokers.

The easiest is to install valkey and use the redis broker.

Using multiple database servers

SQLORM supports using multiple engines.

For example, you can have a primary server for read/writes and a replica for reads.