“We decided to ship “unmanaged Postgres” to buy ourselves time to wait for the right managed Postgres provider to show up. The problem is, fly pg create implies that people are getting a managed Postgres cluster.”
Fly.io lists four months of failures: service discovery, Vault, Postgres, capacity and status pages. Growth jumped from 15% to 30% a month after the Heroku exodus, and capacity planning was his side-hustle spreadsheet. Volumes sit on one physical host, and the docs stopped saying so. Customers ran production on “unmanaged” Postgres that a single command made look managed, and the blog kept cheering the tech stack the whole time.