Thank you for the post. On the “can’t connect to the db because it is not up yet” problem: how about having systemd handle the wait + retry? If the application doesn’t have the wait + retry code, it makes the app code simpler, and its behavior more predictable; it makes the app easier (and quicker) to test. I propose: the app exits when it can’t connect to db, and the systemd service unit is configured for restarting it (`RestartSec=5` and `Restart=on-failure`). Thoughts?