Currently using managed worker pool. Deploying with pool using prefect3 results in the run crashing with "_Reported flow run ... as crashed: Flow run process exited with non-zero status code 1._" Downgrading the pool to use prefect2 kind of solves this crash (also mentioned here: https://github.com/PrefectHQ/prefect/issues/15234#issuecomment-2339776848). But now (with prefect2) it complains about "ModuleNotFoundError: No module named 'prefect_github'" which I need to get GitHubCredentials. Possible to add pip package to the managed work pool, so I added it like this ["prefect[github]"]. Now it runs successfully. But why does the run crash in the first place with the worker pool with prefect3? Had no issues when using .serve instead of .deploy

if __name__ == "__main__":
    main_flow.from_source(
        source=GitRepository(
        url="<https://github.com/me/myrepo.git>",
        credentials=GitHubCredentials.load("github-personal-access-token")
    ),
        entrypoint="prefect/main_flow.py:main_flow",
    ).deploy(
        name="my-deployment",
        work_pool_name="default",
        build=False
    )

I am still learning myself, but .serve is very likely using your "local" environment and has the. necessary packages to run a flow; it sends the logs to Prefect cloud for your to review in the web console Whereas the managed work pool will pull your project and spin up a container for remote execution, which requires the necessary packages to be added at runtime.

Yes that makes sense regarding the missing prefect-github dependency. but I'm not sure if that's the reason for it crashing with "_Flow run process exited with non-zero status code 1."_ instead of giving error with prefect3 worker

I could be wrong but it is likely crashing as the container can’t build or execute without the library, and that’s being captured relative to the flow execution. I think of the managed worker pool as a service that attempts to build an image that executes our code, and in this case, the error is more at the container level. I could be way way off but that’s my mental model at the moment anyway.