How to run a local Prefect server - Prefect

The Prefect CLI is the easiest way to start a local instance of Prefect server.

Start the server

Spin up a self-hosted Prefect server instance UI with the prefect server start CLI command in the terminal:

bash

prefect server start

Open the URL for the Prefect server UI (http://127.0.0.1:4200 by default) in a browser.

Shut down the Prefect server with ctrl + c in the terminal.

Configure the server

Go to your terminal session and run this command to set the API URL to point to a self-hosted Prefect server instance:

bash

prefect config set PREFECT_API_URL="http://127.0.0.1:4200/api"

You can save the API server address in a Prefect profile. Whenever that profile is active, the API endpoint is at that address. See Profiles and configuration for more information on profiles and configurable Prefect settings.

Database configuration

Use SQLite (default)

By default, Prefect uses a SQLite database stored at ~/.prefect/prefect.db. No additional configuration is needed for basic use.

Use PostgreSQL

To use PostgreSQL as your database backend:

Set the database connection URL:

bash

prefect config set PREFECT_API_DATABASE_CONNECTION_URL="postgresql+asyncpg://postgres:yourTopSecretPassword@localhost:5432/prefect"

Start the server:

bash

prefect server start

For more database configuration options, see the database settings reference.

Database management commands

Reset the database

Clear all data and reapply the schema:

bash

prefect server database reset -y

Manage migrations

Apply database migrations:

bash

# Upgrade to the latest version
prefect server database upgrade -y

# Downgrade to the previous version
prefect server database downgrade -y -r -1

# Downgrade to a specific revision
prefect server database downgrade -y -r d20618ce678e

For large databases, you may need to increase the timeout:

bash

export PREFECT_API_DATABASE_TIMEOUT=600
prefect server database upgrade -y

Multi-worker API server

For high-throughput scenarios, you can run the server with multiple worker processes to handle concurrent requests more efficiently:

bash

prefect server start --workers 4

This starts 4 worker processes to handle API and UI requests concurrently.

Requirements for multi-worker mode

Multi-worker mode has specific infrastructure requirements:

PostgreSQL database - SQLite is not supported due to database locking issues
Redis messaging - In-memory messaging doesn't work across processes

Configuration example

bash

# Set PostgreSQL connection
prefect config set PREFECT_API_DATABASE_CONNECTION_URL="postgresql+asyncpg://postgres:password@localhost:5432/prefect"

# Configure Redis messaging
prefect config set PREFECT_SERVER_EVENTS_MESSAGING_CACHE="prefect.server.utilities.messaging.redis"
prefect config set PREFECT_SERVER_EVENTS_MESSAGING_BROKER="prefect.server.utilities.messaging.redis"

# Configure Redis messaging host and port
export PREFECT_REDIS_MESSAGING_HOST="redis"
export PREFECT_REDIS_MESSAGING_PORT="6379"

# Start server with 4 workers
prefect server start --workers 4

<Tip> The number of workers should typically match the number of CPU cores available to your server process, but you may need to experiment to find the optimal value for your workload. </Tip>

Advanced configuration

For advanced deployment scenarios including:

Running behind a reverse proxy
Configuring SSL certificates
Multi-server deployments
Handling migration issues

See How to scale self-hosted Prefect.