09/04/2023, 2:15 AM
Given all the different deployment options now, it would be helpful for the documentation to have examples/use cases/best practices for when to use prefect deploy vs prefect deployment vs flow.serve(). For new users, without a bunch of experimentation, it is unclear and confusing when deciding which to use.
Especially when figuring out the limitations associated with each option when doing a kubernetes/worker/agent based pool. For instance, when deployed to kubernetes using a worker & prefect deploy, it is unclear that the storage block is unused, and that task cache storage becomes broken without additional setup (such as mapping /root/prefect to a shared drive).

Mattias Petter Johansson

09/04/2023, 5:04 AM
On a general note from a user that has just set up a flow: Prefect is amazing but it has optionality sprawl and that meta-developer-experience is not quite managed well in docs an onboarding. In general, I would really love one prefect golden path tbh for us that do it for the first time - one well tested documented line designed to operate on a fresh cloud project and not try to integrate into the production testing ground first, one of google cloud, one for aws, and ane for azure maybe. Then after getting everything your own flows running and having a sense of the whole, one can start understanding the subtleties.

Chris White

09/04/2023, 12:14 PM
Hi Maity - I have some guidance at the bottom of this new doc that I have an open Pr for here: There’s Definitely room for improving our guidance across the board as you both suggest, so the feedback is welcome and appreciated!