It looks like we have a potential design flaw in pulp_deb, that might require reducing the uniqueness constraints on the Release
content model. This would in turn require merging existing colliding content, and then regenerating all other content that references that content via foreign keys (which is numerous) across all existing repository versions. In the cleanest version of this potential fix, I would also be tempted to rename the model from Release
to Distribution
, which would change the API endpoint path. Either way there would be breaking API changes, since I would at the very least drop certain fields from the model.
My question is: Is there any experience of attempting anything like this for any plugin or pulpcore?
Do we already have examples for migrations that reduce a models uniqueness constraints, and then merge the then colliding content?
Do we have examples of migrations that retroactively add altered content to existing repo versions?
I am not yet certain this rather drastic change will ultimately be necessary, but I would like to here everyone’s thoughts regardless, so I can better evaluate the tradeoffs involved.