@Marvin How can I configure a deployment to use a server at a specific IP address?

oops! it seems marvin went on an unexpected hiatus - will investigate

I am using a self hosted server. My intent is to set up the following: • Local dev teams creating Flows on local workstations • Teams push Deployments to S3 (This is already working) • Prefect Server running in AWS getting deployments from S3 ◦ The Prefect Server does not seem to find Deployments in S3, I think I need to set the

PREFECT_API_URL

on local workstations • Prefect Server orchestration Docker Runs on same server? Or Fargate, or ECS (is there a best practice?)

gotcha! you could check out this guide for running an ECS worker to execute deployment flow runs, so • run a worker on ECS as a service (fargate or not is up to you) • you can define a

push

step in your

prefect.yaml

that uploads deployment creators' code at deployment time to S3 • define a

pull

step as well that tells the worker to pull the code from S3 before runtime when you deploy the worker, you'll have to give it the api url to talk to prefect and that ECS service the right role to pull from S3

Ah, the eternal return. I've never left, really. How can I assist you with your fleeting human concerns today?

@Nate What configuration change does a Flow/Deployment contributor need to make to send Deployments to a self-hosted Prefect server instead of Prefect Cloud? Does the developer need to set the PREFECT_API_URL to be the IP/api of the self-hosted server?

in principle, the deployment creator doesn't need to know at all its a matter of the

PREFECT_API_URL

at deployment creation time, which is often from a CI/CD environment that an admin might control a common pattern we see a lot: • flow / deployment contributor / creator checks in a new flow to version control • CI based on version control runs

prefect deploy

with an appropriate

PREFECT_API_URL

in the env