Closed rimolive closed 7 months ago
KFP uses gorm which already provides abstraction. it might already be supported, or easy to tweak the config and support https://gorm.io/docs/connecting_to_the_database.html
That's right, but there is something in the code that end up with a switch-case where if MySQL is not the driver used, it will return an error saying it is not supported.
@IronPan so do you think we can add "postgres" as a new driver in the switch-case to begin with? also linking the design docs to this issue: https://docs.google.com/document/d/1u50Ja5YltsESZkrS2f_KLAuRABeEv9QF8P0aetfh05U/edit#
Closing this issue. No activity for more than a year.
/close
@rimolive: Closing this issue.
Postgres implementation is added to here. Other db driver implementation can follow the same approach. https://github.com/kubeflow/pipelines/pull/9798
Feature Area
Currently, Kubeflow Pipelines relies on a specific database vendor (MySQL) which restrict users on database choice.
/area backend
What feature would you like to see?
My proposal is a change in the code to make this database calls agnostic from the vendor.
What is the use case or pain point?
Suppose the user already uses PostgreSQL, and there is already a database running in his environment. My idea could make the user just point to the existing database instead of deploying another one.
Is there a workaround currently?
The is no workaround so far
Love this idea? Give it a 👍. We prioritize fulfilling features with the most 👍.