ehmatthes
commented
11 months ago Description
Originally, Heroku always provisioned a Postgres database anytime it detected a Django (Python?) app. They recently changed their policy, so I believe you have to explicitly create a Postgres db if you want one. This is similar to Fly's model, I believe.
That presents some challenges:
- [x] In a configuration-only run, how do we detect the available databases?
- [x] If we detect databases, how do we know which one is the right one to use? (My best current thought is to generate a numbered list of the dbs that were found, and ask the user to select which one to use. Maybe state which one we think it should be. Maybe restate the one they chose by name.)
On Fly, I think it was possible to detect the dbs that had been created, but I don't think it was possible to find out much more.
Aim for the happy path first, and don't try to cover all edge cases right away. We're in alpha, declare beta when we're seeing stability, 1.0 after that.
I think Platform.sh is still the one with a single create command, although it's not always easy to install their CLI on Windows.
I don't think I've named this quite so clearly in other issues, but looking back after some time away from this project, one of the most difficult things to deal with has been the need to first create multiple resources on some platforms, and then use those resources together in the same project. It's much easier when a platform has one single
createcommand that creates all necessary resources at once. Failure modes are much simpler, as are recovery efforts. Also, pushing multiple projects is easier on platforms with a singlecreatecommand.