Hi there, is there a database diagram somewhere explaining the structure of flyte metadata for flyteadmin and datacatalog? I'm trying to get a better handle on how flyte uses these and in particular what data is stored in it (with regards to task outputs for example). Are inputs and outputs from tasks ever stored directly in the database or do they reside only in the backend storage (like s3)?

I'm hoping this data management doc would be helpful for you to understand the nuances: https://docs.flyte.org/en/latest/concepts/data_management.html

is there a database diagram somewhere explaining the structure of flyte metadata for flyteadmin and datacatalog?

There isn't one but I recommend you to check out these guides. https://docs.flyte.org/en/latest/concepts/catalog.html & https://docs.flyte.org/en/latest/concepts/admin.html#divedeep-admin

Thank you, so if I understand correctly primitive data types that are returned by tasks are stored in the metadata database, but for large data types or file-like data, only references to locations in blob storage are stored in the database?

Yeah! Only the references are stored.