How-to use caching

[glenthomas]

I am currently looking at ways to reduce GitHub Actions workflow durations and one possible optimisation could be to cache the Terraform CLI, provider plugins and modules.

My understanding is that:

• the Terraform CLI is downloaded from Hashicorp and installed on every run.
• the provider plugins are downloaded from the provider registries and installed on every run.
• the modules are downloaded from their sources on every run.

I am expecting that their may be some performance benefit (maybe a small one) in saving these to the GitHub Actions cache between runs to avoid the multiple downloads that take place.

Workflow caching for Terraform is not something that I have attempted before...

Does anyone have experience of implementing any of this with the dflook actions?
Does anyone have any opinions on why it might not be a good idea or not be possible?

[dflook]

The actions try and cache the terraform binary and plugins on disk, but if you are using github hosted runners (or ephemeral self-hosted runners) the cache will start empty for each job.

There was previous discussion in #105. Essentially, If you know the path to the cache on disk you can use it with actions/cache to save and restore those paths and carry the cache between jobs.

Today, On github hosted runners you should be able to cache terraform plugins using the path:
${{ runner.temp }}/_github_home/.terraform.d/plugin-cache
and terraform binaries using
${{ runner.temp }}/_github_home/.dflook-terraform-github-actions

This shouldn't be two paths - in the future everything will be moved under the second path. I haven't actually tested this myself so it would be great to see a working example.