We are pleased to see the excitement, feedback, and positive energy about deploying your flows to Prefect 2.0 and building a strategy for reliable CI/CD pipelines.
Deployments have been redesigned fairly recently, and the engineering team invested their energy into the GA release that happened not even a week ago (!). Therefore, we haven’t yet had enough time to build best practice deployment recipes with packaging dependencies and CI/CD workflows. We understand this is required for you to reliably migrate over to Prefect 2.0, and we will work with our full energy to provide that in the coming weeks.
Given that we are actively working on further improving the deployment UX and making it more transparent how to apply it to various use cases, deployment patterns, and execution environments, we would recommend waiting for official recipes coming in the following days and weeks before committing to a production deployment story just yet. There are plenty of things you can already do in migrating your workflows, such as migrating your secrets and configuration data using blocks, experimenting with the new task result handling and states, and parallel execution using mapping.
We’ll continue shipping new releases that will further improve the deployment experience and CI/CD workflows - you can already start thinking about deployment patterns for your execution layer (agents and work queues) and team organization (workspaces, tags), but we’d encourage you to wait before committing to deployment patterns for flows and packaging code dependencies.
If you want to give us feedback about your experience with Prefect 2.0, we encourage you to articulate the problems that you try to solve rather than focusing on features that existed in the beta version (such as, e.g., packagers or flow runners) but which no longer exist in the General Availability release. While we can’t guarantee backward compatibility with beta releases, we will almost certainly be able to solve the problems that you raise.
Thanks, and happy engineering!