Hi, I’m trying to deploy a trivial flow in Prefect 2.0 using S3 and the RemoteFileSystem storage block. The block is properly registered with Orion. For some reason

prefect deployment build

gives me a deployment YAML, but is no longer producing the manifest JSON file. It’s only producing the deployment YAML. How can I debug this? Here is the build command used:

prefect deployment build myflow.py:myflow -n myflow-deployment -t dev -ib kubernetes-job/k8dev -sb remote-file-system/dev

In 2.0.4 the manifest.json is no longer used. The information now comes directly from the deployment YAML. We simplified deployment process by just using the single file. 🙂

Ah, so my local orion and agent are running the image

prefecthq/prefect:2.0.3-python3.10

. I guess that explains why my local k8s cluster started throwing this when I ran

prefect deployment apply

prefect.exceptions.PrefectHTTPStatusError: Client error '422 Unprocessable Entity' for url '<http://localhost:4200/api/deployments/>'
Response: {'exception_message': 'Invalid request received.', 'exception_detail': [{'loc': ['body', 'manifest_path'], 'msg': 'none is not an allowed value', ...

I talked to the engineering team and learned that we only guarantee compatibility of equal versions for self-hosted servers, but we generally strive for compatibility between different versions.

how about on the cloud side? I run self-hosted during development, but cloud during production.

We intend to have backwards compatibility between the current Cloud version and a number of previous Prefect releases. The team is thinking about the issue and how to communicate if/when there are breaking changes.