Hi! Can someone tell me if I can use

pull: prefect.deployments.steps.git_clone:

with Prefect flows running in a Kubernetes job? I'm exploring ways to speed up execution. Currently, whenever I update my flow code, I have to rebuild the image used in the Kubernetes job each time.

you can absolutely use the

git_clone

pull step for kubernetes jobs. the

entrypoint

is the path to your flow function relative to the working directory (set either in your dockerfile or another pull step, or

/opt/prefect

if you're using a prefect base image and don't change the working directory). prefect will supply its own container startup command by default for running a flow, which will then use the entrypoint to execute your flow function, so you don't need to provide that kind of information in your dockerfile

I've played with the configuration a little bit. Here's what I've set up: prefect.yaml:

deployments:

- name: test-dep

path: /test-dir

entrypoint: test.py:sleep

work_pool:

name: my_k8s_work_pool

pull:

- prefect.deployments.steps.set_working_directory:

directory: /app

- prefect.deployments.steps.git_clone:

repository: <https://github.com/my_git_user/my_git_project.git>

dockerfile:

FROM prefecthq/prefect:2-python3.11-kubernetes

RUN mkdir /app

COPY test.py /app/test.py

From the execution logs, it seems that Prefect is searching for the flow function in

/app/my_git_project/test.py

. This indicates that it first enters the working directory

/app

and then navigates to the cloned

my_git_project

directory. It appears that the

path: /test-dir

directive is ignored in this case. Am I right?