@Marvin when Prefect is deployed on one Linux server, is it possible to launch flow on another Linux server, e.g. to run specified docker container there?

To run a Prefect flow on a different server, including running a specified Docker container, you'll need to: 1. Deploy the Flow to Prefect Cloud: To allow the flow to be orchestrated from anywhere. 2. Use Work Pools/Workers: Configure the workers to run on different servers, targeted using work pools. ### Step-by-Step Guide 1. Deploy Your Flow: You can deploy your Prefect flow using the

deploy()

method:

python
from prefect import flow
from prefect.deployments import Deployment

@flow
def my_flow():
    print("Running my flow")

if __name__ == "__main__":
    Deployment(name="my_deployment", flow=my_flow).apply()

2. Setup Work Pools in Prefect Cloud: Configure work pools on the servers where the flow should run. This can be done in the Prefect Cloud UI under the "Work Pools" section. 3. Run Workers: On the target server where you want to run the flow, start a Prefect worker from the specified work pool:

sh
prefect worker start --pool your_work_pool_name

4. Specify Docker Container: If you need the flow to run in a specific Docker container, specify the image in your deployment configuration:

python
Deployment(name="my_deployment", flow=my_flow, work_pool_name="your_work_pool_name", image="your-docker-image").apply()

### Links for Reference: - Deployment Configuration - Work Pools This setup will ensure your flow can be orchestrated by Prefect but executed on another server, potentially using a specified Docker container.

@Marvin consider only on-premises Prefect deployment, don't consider Prefect Cloud.

For on-premises Prefect deployments, you can configure the setup to run flows on different servers by following these steps: 1. Deploy the Flow: Use the deployment capabilities of Prefect. 2. Configure Workers on Target Servers: Ensure you have Prefect workers running on the servers where you want the flows to execute. Here is a detailed approach: ### 1. Deploy the Flow First, create a deployment for your flow. This is similar to the previous example but ensuring it is meant for on-premises:

python
from prefect import flow
from prefect.deployments import Deployment

@flow
def my_flow():
    print("Running my flow")

if __name__ == "__main__":
    Deployment(name="my_local_deployment", flow=my_flow).apply()

### 2. Start Workers on Target Servers On the server where you want to run the flow, start a Prefect worker:

sh
prefect worker start --pool work-pool-name

This command ensures that the worker is listening for tasks assigned to it from the specified work pool. ### 3. Specify Docker Container (If Needed) If the execution needs to be within a Docker container, modify the deployment to include the Docker image:

python
Deployment(
    name="my_local_docker_deployment",
    flow=my_flow,
    work_pool_name="your_work_pool_name",
    image="your-docker-image"
).apply()

### Example Code for Reference Here is some example code to get you started on running tasks remotely, which can be configured similarly for deploying and running flows: LINK: https://raw.githubusercontent.com/zzstoatzz/prefect-code-examples/main/examples/tasks/submitting.py EXAMPLE:

python
import time
from prefect import flow, task

@task
def sleep(t: int):
    print(f"Going to sleep for {t} seconds...")
    time.sleep(t)
    print("...and I'm awake!")

@flow(log_prints=True)
def my_flow():
    sleep(3)
    future = sleep.submit(2)
    future.wait()
    sleep.map([5] * 10)

if __name__ == "__main__":
    my_flow()

### Additional Resources - Deployments Documentation - Workers Documentation By following these steps, you can configure Prefect in an on-premises environment to deploy and run flows on different servers.

The Prefect server and workers are independent from each other, so as long as the worker can reach the server for requesting work over the network, the answer is yes

OK, thanks, now I understand that it's enough to have one Prefect server (UI) on one machine, and Prefect workers on all machines where flows are needed to run. Though I don't yet understand how to communicate with workers on other machines (does every worker have its API url?...), but I will check it later...

the API url is the server, so each worker will use the same one to find out from the server when there are flows to run

Oh, you mean that worker needs to see environment variable with API url to server?

correct!