Hey team, I have been using Prefect 2.13.8 and my main flow is run via the

run_deployment

command. The entrypoint flow

A

calls a series of concurrent subflows

B

. Both

A

and

B

flows have similar

flow

decorator definitions, including retry configurations. I've noticed that whenever the parent flow

A

fails, it seems to automatically retry; however, if sublow

B

fails, it runs into an issue with the error:

ValueError: Path /root/.prefect/storage/e1f52b58ca3c4687a39c224g5cbb1223 does not exist.

I believe this may be due to something regarding my docker / k8 setup. Is this a known problem? I would love to get some tips on how to resolve this 🙂

Hey @Gabe Villasana! When a flow retries, it attempts to load results for the tasks runs in the previous flow run. It looks like your flow is persisting results to the local filesystem, which is the default behavior. When you’re using

run_deployment

in conjunction with a k8s setup, the flow retry is likely running on a different machine and can’t find those persisted results. You can solve this by writing your results to a remote location like S3 to access results across flow runs. This section of the docs contains information about changing where your results are stored. It’ll involve configuring a block that points to a remote storage location. Let me know if you have further questions about setting that up!

This makes sense. I'll give this a try now and lyk if I have further questions. Thank you 😄