minerva icon indicating copy to clipboard operation
minerva copied to clipboard

Removing S3 folder does not delete dataset that references it

Open kotfic opened this issue 9 years ago • 2 comments

Datasets reference items in the s3 folder, removing the s3 folder does not cascade a delete of datasets that reference it.

kotfic avatar Aug 20 '15 13:08 kotfic

I don't fully understand what is desired here. Let's say we have a dataset that is derived from an item in S3 but no longer relies on the original item. Would you want this dataset deleted?

mgrauer avatar Aug 20 '15 13:08 mgrauer

No I don't think we would want to delete it in that case,

The scenario I was thinking of is this:

  • Add an s3 dataset through the minerva interface
    • this creates an s3 assetstore
    • this creates a folder under the User Name > minerva > s3 folder where the item data is imported based on the s3 bucket name (e.g. nasanex) and the prefix (e.g., NEX-DCP30)
    • this creates an item under User Name > minerva > datasets (e.g. NEX Test)
  • Someone goes into the backend in girder and removes the User Name > minerva > s3 > nasanex > NEX-DCP30 folder.
  • Now the NEX Test Item has metadata that refers to a folder that no longer exists.
  • On the front end if you go to browse NEX Test the widget shows an empty folder.

Ideally we would delete dataset (or "data source") items on s3 folder removal.

kotfic avatar Aug 20 '15 13:08 kotfic