Governing old projects : easing the archive or deletion

dupontni

Hi,

we have hundreds of dataiku projects on our instances, many of them are just old versions or failed experiments and are wondering if someone already worked on this.

Today we are listing projects through the Python API, exposing to end users a list of projet keys, associated with the owner and lastModifiedOn timestamp (which shoul be completed with the lastactivity timestamp : no modification doesn't mean no activity) : the list is provided through a google sheets to help them clean the old projects manually

We are wondering how a plugin could help our end user to get more self service

• Plugin Settings would let us set a timeframe to isolate "old" projects based on lastModification and/or lastActivity, providing a conditional formatting to help processing the list
• The list could be displayed to the end user directly in dataiku, providing them a link to trigger Archiving (i.e. set project status to archive) or deletion. Projects having a owner which account has been disabled could be a hint too

We didn't find any initiative in the community relative to old/duplicate project management and would like to know if someone in the community already encountered the issue and would be ready to describe how you fix it ?

Operating system used: RHEL

Marlan

We just recently did something similar to what you describe (listing projects with info via the API) as part of an upgrade of projects to a later Python version. A tool that provides a UI for self-service would be nice. I'd think that this could be a good use case for a web app.

Marlan

Turribeach

So along the same lines we are working in a process that will inform users of projects that will be tagged as inactive. This will be be done via the Python API and we will email relevant users using send mail steps in a Dataiku scenario. There is no reason why you can't do the same. You could also be clever and email the project owner, any users with edit permisions in the project or the last modifier. All of the APIs are there, just need some coding. And like @Marlan says this could be exposed via a Dataiku Webapp if you wanted a GUI where they could perform the actions in bulk or against individual projects.