Packaging a recipe is basically knowing a bag of about 20 tricks. Once you learn them, there is nothing more to learn. It can seem daunting at first but it's simple and easy to do once you know the tricks.
The nice thing about packaging is that only one person has to do it and then we all benefit. We've seen that over time, the core of the configuration doesn't really change. New options and versions might come but the config remains quite stable. This is good since it means that your packaging work stays relevant and useful for other maintainers & operators as time goes on.
Depending on your familiarity with recipes, it might be worth reading [how a recipe is structured](/maintainers/handbook/#how-is-a-recipe-structured) and making clear you understand [what a recipe is](/glossary/#recipe) before continuing.
The ideal scenario is when the upstream project provides both the packaged image and a compose configuration which we can build from. If you're in luck, you'll typically find a `Dockerfile` and a `docker-compose.yml` file in the root of the upstream Git repository for the app.
Let's take a practical example, [Matomo web analytics](https://matomo.org/). We'll be making a Docker "swarm-mode" `compose.yml` file.
Luckily, Matomo already has an example compose file in their repository. Like a lot of compose files, it's intended for use with `docker-compose`, instead of "swarm mode", but it should be a good start.
Open the `compose.yml` in your favourite editor and have a gander 🦢. There are a few things we're looking for, but some immediate changes could be:
1. Let's bump the version to `3.8`, to make sure we can use all the latest swarm coolness.
2. We load environment variables separately via [`abra`](/abra/), so we'll strip out `env_file`.
3. The `/var/www/html` volume definition on L21 is a bit overzealous; it means a copy of Matomo will be stored separately per app instance, which is a waste of space in most cases. We'll narrow it down according to the documentation. The developers have been nice enough to suggest `logs` and `config` volumes instead, which is a decent start.
4. The MySQL passwords are sent as variables which is fine for basic use, but if we replace them with Docker secrets we can keep them out of our env files if we want to publish those more widely.
5. The MariaDB service doesn't need to be exposed to the internet, so we can define an `internal` network for it to communicate with Matomo.
6. Lastly, we want to use `deploy.labels` and remove the `ports:` definition, to tell Traefik to forward requests to Matomo based on hostname and generate an SSL certificate.
The resulting `compose.yml` is available [here](https://git.autonomic.zone/coop-cloud/matomo/src/branch/main/compose.yml).
The rest of this guide assumes you have a Co-op Cloud server going -- we'll use `swarm.example.com`, but replace it with your own server address. Head over to [the operators tutorial](/operators/tutorial) if you need help setting one up.
Then, open the `DOMAIN` you configured (you might need to wait a while for Traefik to generate SSL certificates) to finish the set-up. Luckily, this container is (mostly) configurable via environment variables, if we want to auto-generate the configuration we can use a `config` and / or a custom `entrypoint` (see [`coop-cloud/mediawiki`](https://git.autonomic.zone/coop-cloud/mediawiki) for examples of both).