Need to apply when no changes detected on plan?

[seboudry]

Hi! And happy new year ;)

For now, when drift detection is configured to run on a schedule, a plan is run regularly.
Even if the no changes is detected, the plan-sum hash changes.
So a new apply run is started because hash is different from apply-sum hash.

So, I raise the question about the need to run an apply when a plan detect no changes.
This create "empty" apply logs on S3 and waste compute resources.

[AlanLonguet]

Hi !

Maybe I'm mistaken but I think you can get changes to outputs without having any effective changes in the resources it deploys.

It's a behavior we already noticed but didn't change it, we might be able make this an opt-out feature though

[seboudry]

When you talk about outputs it's burrito files uploaded to datastore or TF outputs?

But I still don't get what you thinking about, forgive me.
Do you have an example or an explanation?

[AlanLonguet]

TF outputs, I think you can have an "empty" plan (meaning no resources changed) but you changed the format of one your outputs and this needs to be applied so the outputs are saved in the state.

This might not be of use for vanilla terraform but for terragrunt which relies heavily around it I think we can't really bypass this empty apply.

However we can be smarter in the way we determine if we need to apply by checking if there's changes to outputs and/or changes to resources.

WDYT @LucasMrqes @corrieriluca ?