Terraform registry caching

[ilithanos]

One caveat with the tofu-controller setup when running it in cloud managed k8s clusters is the cost on data transfers.

I've been looking around for options on solving this, for modules themselves it's mostly solved by using OCI packaged modules. But since we are pulling providers etc on runners every time they run, using a provider mirror with pull through cache could save a lot on the cost of running the controller.

At my current position we have been looking into using the boring-registry for this, but are having some issues getting it to behave.

I'm unsure if the caching part of the implementation for allowing safe deletion of modules even if their sources are removed could also solve this issue. But if it can't i see 2 possible options for us as a community to solve this:

1. Implement caching in some form as part of the tofu-controller

• Pros: simple to use from a user perspective.
• Cons: added complexity within the tofu-controller

2. Create a seperate project under the flux-iac organization for building a provider registry that works with the tofu-controller

• Pros:
  • Seperate project will avoid adding the complexity to the controller
  • Could be used by other tools than the tofu-controller
• Cons:
  • More complex setup when first setting it up
  • Another project to find maintainers for.

What does the rest of the community think?