Newly refactored module

[js-timbirkett]

Hello 👋

Firstly, this module is great and has a lot to deal with as standing up an EKS cluster and nodes / workers is not a trivial task.

There have been issues regarding orchestrating the creation of the control plane and the workers / nodes [#519], issues questioning complexity or coupling within the module [#635], questions with regards to LC / LT worker_groups and whether they should be merged or simplified [#563], and issues around numeric indexes and modifying arrays in general.

I had a go at creating a new, refactored version of this module and my attempt is here:
https://github.com/devopsmakers/terraform-aws-eks

It breaks things out into sub-modules that can be used by both the parent module (to maintain the simple single module gets you a working cluster) and directly by the user of the module for more complex situations (Custom CNI).

It also allows a map of maps for worker groups and the code has been simplified a bit (reduction of the death by lookups in the current worker related code).

I made a few tough decisions that might stimulate some debate and possibly some outrage. These are highlighted in the repos README.

I'm not saying it's perfect (yet), but I think it would make a good base for a fairly major refactor of this module and I'd be glad to assist in any way I can. Feel free to use any ideas or the codebase itself to further improve this module.

Thanks!

[barryib]

Hello @js-timbirkett thanks for opening this discussion and for your work. It's a good start.

Can you please open a PR so we can iterate and give feedbacks ?

For the design decisions:

• To me, we can move worker groups in a separate submodule like we already did for managed node groups.
• As for dropping LC support, I don't have a strong opinion on this. I personally don't use LC at all, but I don't know if there are lot of people out there which still rely on it. See Is the complexity of this module getting too high? #635 (comment)

@max-rocket-internet @dpiddockcmp @antonbabenko please advice (mostly for the LC removal)

[antonbabenko]

+1 for dropping LC. I think we all agreed in some previous issues that dropping support for LC is the way to go.

I wonder, are there any downsides of using LT at all comparing to LC?

[max-rocket-internet]

I wonder, are there any downsides of using LT at all comparing to LC?

No downsides to LT that I know of, LC is just older and has less features.

[max-rocket-internet]

I had a go at creating a new, refactored version of this module

Sounds great but we need a PRs to review and a path forward for current users.

I look forward to seeing some PRs 😃

[js-timbirkett]

Hey 👋

The refactored module is still very much in a beta state, when I've finished shaving the Terraform yak and double checking outputs, inputs etc... I'll open a big ole PR for further discussion and improvements.

It'll definitely be a major release and not something that's easily migrate-able to without a lot of Terraform state moving or importing (could probably write up a guide on that).

[chancez]

@js-timbirkett I'll be giving your worker group module a try. I'm vendoring it though since I need to change ignore_changes for the ASG to not ignore desired capacity since we're using TF, not cluster autoscaler for managing workers.

[js-timbirkett]

Oh nice @chancez - Thanks for the 🐛 fix!

I'll be doing some final editing of my version of the module early next week when I have some time off work. Then I'll open what will be a rather large PR against this module and see how it goes...

[barryib]

@js-timbirkett any update here ?

[barryib]

@js-timbirkett are you still working on this ?