Trying to get our first prefect deployment set up. We have a single parent flow that is bundled with some other code we want to host on an EC2 instance. In deploying via GitHub Action, I am wondering the best way to handle different processes that are required - i.e. I have a python file with

my_flow.serve(...)

and then I also need to make sure the prefect agent is running

prefect agent start -q default

and that the server is running

prefect server start

. Is there a simple way to accomplish this in a deployment script? Should I set these up as separate actions? Or should I look at how to accomplish them with something like

multiprocessing

?

if you are serving your flow, you don't need an agent - serving makes the flow runnable straight from the UI

thanks. Tried this, I can get my flow to deploy to my EC2 instance, and I can now see the deployment in the cloud UI. But the

my_flow.serve(...)

command is holding the GitHub Action open. What's the best way to send a deployment to long-lived infrastructure without keeping the deployment action open?

serve

is similar to starting a worker/agent but specific to the flows you're serving. for deploying to somewhere there's a worker running, then you'll want to use

deploy

Trying that, seems that using

deploy

requires a worker to be started, which is a long running process. Looking at starting the worker on EC2 startup - is there a better way to do this?

we typically see people starting their worker with the helm chart in k8s, or using one of our docker images with the

prefect worker start --pool <your-work-pool-name>

as the startup command on whichever computer service they're using