<@ULVA73B9P> What's the deal with `ephemeral-prefe...
# ask-marvin
r
@Marvin What's the deal with
ephemeral-prefect/api
?
m
The term "ephemeral-prefect/api" seems to appear in issues related to server and logging errors in the PrefectHQ/prefect repository. Here are some details: 1. Internal Server Error during prefect-dbt tests: - Users encountered a
500 Internal Server Error
when running certain tests. - The error logs indicate issues with the API endpoint for logging (`http://ephemeral-prefect/api/logs/`). - Source: GitHub Issue #14297 2. Functionality degradation in Prefect 3.0.0rc1: - Users reported that submissions of flows in a Flask app (served via gunicorn + gevent) worked only on the first attempt in Prefect 3.0.0rc1; subsequent attempts failed. - This error might be by-passed when using the Flask dev server (
python app.py
). - Source: GitHub Issue #13744 3.
on_cancellation
Hook Not Triggered in Prefect 2.16.4
: - The
on_cancellation
hook was not triggered for canceled flows, despite the workflow appearing as canceled in the UI and client. - Logs indicated an issue with the hook execution. - Source: GitHub Issue #13834 Based on these issues, the
ephemeral-prefect/api
endpoint is associated with server-side errors and logging issues in certain testing and operational scenarios. The term is likely related to ephemeral instances of the Prefect API server used in these contexts. It would be best to monitor or address the specific implementation details around this endpoint in your setup. For further details and potential solutions, you can explore the linked issues directly. If you need more help, please let me know how I can assist further!
n
@Robert in general the ephemeral api is just an instance of the fastapi app that runs "when you need it" so if you
pip install prefect
and do nothing else and then run some flows, we'll spin up an ephemeral instance of the API so all that run data gets persisted to the db and stuff its meant for an easy on ramp before engaging with the conceptual bulk of our OSS server or cloud
r
Thanks Nate! I think what happened is I neglected to set the
PREFECT_API_URL
env var
n
👍
179 Views