Design CI/CD deployment and remote hosting architecture

[onecrayon]

I need server architecture designed with a related CI/CD flow, but I hate doing this stuff myself. Things I would love to have:

• [x] Recommendations for what architecture to use (Docker Swarm deployed to one or more Linode VPSes; e.g. DockerSwarm.rocks but simpler? Bare metal VPS for API with managed PostgreSQL database--e.g. Heroku or DigitalOcean? Something else?)
• [x] CD workflow that preferably integrates with GitHub Actions to automates PR testing and deployments to remote architecture upon merging to main branch

My goals for remote architecture are: low maintenance (I don't want to worry about setting up database backups, for instance; they just need to happen); ideally define infrastructure as code, since I'd rather not manage stuff through control panels; automated deployments (which means deploying latest code, automatically updating dependencies based on the Poetry lock file, and running any pending migrations); and ideally scalable infrastructure (things like Digital Ocean and Linode look very nice, since I can start very small and dial up the service as needed; Heroku similarly looks great, except their Postgres options are $$$).

[onecrayon]

Current plan:

• Caddy runs as a reverse proxy in front of Gunicorn, which handles process management using the Uvicorn worker class for ASGI server functionality
• Database is hosted on a separate server (preferably a managed instance)

Open questions:

• [x] Figure out if the Caddy setup works properly for Docker Swarms! I'd really like to be able to deploy this as a swarm, even if it only has one node to start. That way I can scale horizontally quite easily.
• [x] Document every step that needs to be taken to deploy, but maybe not in the README. Actually, definitely not here in the README, because I'm the only one who will ever be deploying this code.
• [x] Figure out if there's a package registry I can use (probably necessary for CD)
• [x] Figure out how to handle .env creation for CD
• [x] Figure out how to handle migrations in a CD workflow
• [x] Ensure that Caddy volumes are hooked up to persistent storage of some sort! That stuff can't be ephemeral

[onecrayon]

Found my perfect solution! https://render.com supports everything I care about (standalone, managed database; arbitrarily scalable Docker nodes for the API; static site hosting for the Vue.js app; automated builds when pushing to the main GitHub branch).

Only thing I need now is a CI workflow for automated testing in GitHub PRs.