Dylancyclone

No that's totally fair! I'm a huge fan of making things reproducible since I've ran into too many situations where things need to be rebuilt, and always open to ways to improve it. At home I use ansible to configure everything, and at work we use ansible and declare our entire Jenkins instance as (real) code. I don't have the time (and I'm low-key scared of the rabbit hole that is) Nix, and to me my homelab is something that is configured (idempotently) rather than something I wanted to handle with scripts.

I even wrote some pytest-like scripts to test the playbooks to give more productive errors than their example errors, since I too know that pain well :D

That said, I've never heard of PyInfra, and am definitely interested in learning more and checking out that talk. Do you know if the talk will be recorded? I'm not sure I can watch it live.

I love the "Warning: This talk may cause uncontrollable urges to refactor all your Ansible playbooks" lol I'm ready

While it is true that Ansible is a different tool that you need to learn the basics of (if you want to edit/add applications), all of the docker stuff is pretty comparable. For example, this is the equivalent of a docker compose file for SilverBullet (note taking app): https://github.com/Dylancyclone/ansible-homelab-orchestration/blob/main/roles/silverbullet/tasks/main.yml

You can see it's volumes, environment variables, ports, labels, etc just like a regular docker compose (just in a slightly different format, like environment variables are listed as env instead of environment), but the most important thing is that everything is filled in with variables. So for SilverBullet, any of these variables can be overwritten, and you'd never have to look at/tweak the "docker compose." Then, if any issue is found in the playbook, anyone can pull in the changes and get the fix without any work from themselves, and if manual intervention is needed (like an app updated and now requires a new key or something), the playbook can let you know to avoid breaking something: https://dylancyclone.github.io/ansible-homelab-orchestration/guides/updating/#handling-breaking-changes

If you'll let me self promote for a second, this was part of the inspiration for my Ansible Homelab Orchestration project. After dealing with a lot of those projects that practically force you to read through the code to get a working environment, I wanted a way to reproducably spin up my entire homelab should I need to move computers or if my computer dies (both of which have happened, and having a setup like this helped tremendously). So far the ansible playbook supports 117 applications, most of which can be enabled with a single configuration line:

immich_enabled: true
nextcloud_enabled: true

And it will orchestrate all the containers, networks, directories, etc for you with reasonable defaults. All of which can be overwritten, for example to enable extra features like hardware acceleration:

immich_hardware_acceleration: "-cuda"

Or to automatically get a letsencrypt cert and expose the application on a subdomain to the outside world:

immich_available_externally: true

It also comes with scripts and tests to help add your own applications and ensure they work properly

I also spent a lot of time writing the documentation so no one else had to suffer through some of the more complicated applications haha (link)

If you'll let me self promote for a second, this was part of the inspiration for my Ansible Homelab Orchestration project. After dealing with a lot of those projects that practically force you to read through the code to get a working environment, I wanted a way to reproducably spin up my entire homelab should I need to move computers or if my computer dies (both of which have happened. So far the ansible playbook supports 117 applications, most of which can be enabled with a single configuration line:

<application_name>_enabled: true

For example:

portainer_enabled: true
nextcloud_enabled: true

That is very true, I suppose a more accurate way to say it would be the playbook does not need to run as the root user, and can instead use the permissions of a regular user. This lets all the volume mounts be owned by your user, instead of root. I think it's still an important distinction to make though, since by not running the playbook as root, the playbook can't directly change any of your server's settings, and only has the access the user you're SSHing as has.

Yes, this playbook is intended to orchestrate an already set up environment. I know Ansible can easily install and set up docker (using something like the awesome https://github.com/geerlingguy/ansible-role-docker), but I decided against it (at least for now) for two main reasons: Firstly to avoid becoming the root user, and secondly to avoid the Ansible role installing a second version of Docker, causing things to break. I ran into that myself while testing this playbook, where I had set up a Ubuntu VM, told the installer to install Docker during setup not knowing it would install it through snap, then the Ansible playbook would install docker again through conventional means causing a lot of strange problems. So instead I opted to let the user install docker however they'd like and not have any gotchas like "Remember to add --skip-tags="docker" if you installed Docker during OS installation on Ubuntu" or uninstalling their version of docker for them

This is definitely inspired by ansible-nas! I'd also used it for awhile, and made my own fork to add/fix things since the project has gone a little dormant. I started making so many changes though that I started fresh and it turned it into a whole project of it's own. You can see a list of differences here: https://dylancyclone.github.io/ansible-homelab-orchestration/guides/introduction/

This project was basically how I learned Ansible haha. It is incredibly powerful and this really only scratches the surface