Hi Prefect community!
I have a question about orchestrating workflows using Prefect when there are steps within the workflow that have different infrastructure requirements.
For example if I have a workflow I would like to orchestrate using Prefect that has these steps
(DBT job on Snowflake) -> (ML training job).
It would be great to provision a small Fargate task with a Docker image with just the DBT project for the first one, and then provision an ECS task with significantly larger resources and a different Docker image for the ML training job.
Is there a way to do this by creating dependencies between Prefect deployments or by specifying infrastructure at the task or sub-flow level?
✅ 1
Ethan Homan
11/02/2022, 7:12 PM
Also for more context, I am very new to Prefect and have been trying it out using ECSTasks with the integrated storage + infrastructure pattern on the Docker images.
k
Kevin Grismore
11/02/2022, 7:13 PM
I highly recommend checking out
run_deployment
(imported like:
from prefect.deployments import run_deployment
). You can deploy flows that need their own infrastructure, like docker images, and then run those deployments from another flow.
Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.