Prefect Cloud can run your flows on your behalf with Prefect Managed work pools. Flows that run with this work pool do not require a worker or cloud provider account—Prefect handles the infrastructure and code execution for you. Managed execution is a great option for users who want to get started quickly, with no infrastructure setup.

​
Create a managed deployment

  1. Create a new work pool of type Prefect Managed in the UI or the CLI. Use this command to create a new work pool using the CLI: 
    prefect work-pool create my-managed-pool --type prefect:managed
  2. Create a deployment using the flow deploy method or prefect.yaml. Specify the name of your managed work pool, as shown in this example that uses the deploy method:
    managed-execution.py
    from prefect import flow

if __name__ == "__main__":
    flow.from_source(
        source="https://github.com/prefecthq/demo.git",
        entrypoint="flow.py:my_flow",
    ).deploy(
        name="test-managed-flow",
        work_pool_name="my-managed-pool",
    )
  3. With your CLI authenticated to your Prefect Cloud workspace, run the script to create your deployment: 
    python managed-execution.py
  4. Run the deployment from the UI or from the CLI. This process runs a flow on remote infrastructure without any infrastructure setup, starting a worker, or requiring a cloud provider account.

​
Add dependencies

Prefect can install Python packages in the container that runs your flow at runtime. Specify these dependencies in the Pip Packages field in the UI, or by configuring job_variables={"pip_packages": ["pandas", "prefect-aws"]} in your deployment creation like this: 
from prefect import flow

if __name__ == "__main__":
    flow.from_source(
        source="https://github.com/prefecthq/demo.git",
        entrypoint="flow.py:my_flow",
    ).deploy(
        name="test-managed-flow",
        work_pool_name="my-managed-pool",
        job_variables={"pip_packages": ["pandas", "prefect-aws"]}
    )
Alternatively, you can create a requirements.txt file and reference it in your prefect.yaml pull step.

​
Networking

​
Static Outbound IP addresses

Flows running on Prefect Managed infrastructure can be assigned static IP addresses on outbound traffic, which will pose as the source address for requests interacting with your system or database. Include these source addresses in your firewall rules to explicitly allow flows to communicate with your environment.
  • 184.73.85.134
  • 52.4.218.198
  • 44.217.117.74

​
Images

Managed execution requires that you run an official Prefect Docker image, such as prefecthq/prefect:3-latest. However, as noted above, you can install Python package dependencies at runtime. If you need to use your own image, we recommend using another type of work pool.

​
Code storage

You must store flow code in an accessible remote location. Prefect supports git-based cloud providers such as GitHub, Bitbucket, or GitLab. Remote block-based storage is also supported, so S3, GCS, and Azure Blob are additional code storage options.

​
Resources & Limitations

All flow runs receive:
  • 4vCPUs
  • 16GB of RAM
  • 128GB of ephemeral storage
Maximum flow run time is limited to 24 hours. Every account has a compute usage limit per workspace that resets monthly. Compute used is defined as the duration between compute startup and teardown, rounded up to the nearest minute. This is approximately, although not exactly, the duration of a flow run going from PENDING to COMPLETED.

​
Next steps

Read more about creating deployments in Run flows in Docker containers. For more control over your infrastructure, such as the ability to run custom Docker images, serverless push work pools are a good option.