Hi, I’m wondering if it’s possible if subflow run on different architecture? For example, if I have • master-flow which trigger subflow-1 and subflow-2 • subflow-1 needs more CPU & memory -> package in a deployment with more CPU & memory in the worker config for ECS task • subflow-2 needs less CPU & memory -> package in a deployment with different (and less) amount of CPU & memory in the worker config for ECS task If it’s possible, will it be done through deployment trigger or simply trigger the subflows in the master-flow after created the deployments for subflows?

You cant use different infra for direct subflow call AFAIK.

If the subflows are in line with your main flow (same file or same thread) as @Deceivious mentioned, you can’t split up a running flow to different infra mid-flow. You can however register each as their own deployment entirely, and call

run_deployments

as mentioned - this will ensure each gets its own process / job; they would run with whatever the work-pool has, or you could override the job params in

infra_overrides

when you defined the deployment

Thank you, both. I could make it to try the run. Anw, I could see that if the flow run in

run_deployment

failed, the

master-flow

run still complete successfully. Is there a simple way to mark the

master-flow

run to failed when

run_deployment

failed?

run_deployment

returns a flow run id. You can check the state of the flow run after it completes. Check for documentation on

timeout

in

run_deployment

. And

FlowRun

in prefect.client.schema

Big thanks to you, @Deceivious!