Hi… It shouldn’t actually be necessary to modify fly.toml for this. That’s because flyctl can bridge the MariaDB Machine’s port over temporarily, so it behaves as if it’s listening at localhost:13306 on your development laptop/desktop. (What’s really happening is that flyctl is shuttling packets local ↔ remote behind the scenes.) Alternatively, you can run the dump command on the Machine itself, and then download the result via SFTP.
I admittedly don’t know MySQL itself that well, but here’s how that looks with Postgres:
https://community.fly.io/t/pg-dump-always-gets-interrupted-while-backing-up-postgres-db/23842
And here are some MySQL-specific tips for the bridging way:
https://fly.io/docs/app-guides/mysql-on-fly/#access-the-database-from-outside
(Note that you leave flyctl proxy running for as long as you need the connection. This surprises many people.)
Finally, separating the web-app from the database has many advantages and one near-necessity on the Fly.io platform: single-volume deployments are unusually risky here. If you’re taking regular backups and can afford to lose all the writes that occurred in the interim, then you may be ok(-ish), but many people have been unpleasantly surprised to find that they’ve lost entire databases, permanently…