Thank you very much for your module, I'm using it here and it works well.
I'm not sure if you could call it a bug but based on my experience with the module I have discovered that Terraform resets the number of worker nodes to node_count when a subsequent run is triggered. This causes unnecessary scalings as the applications deployed on the cluster will immediately trigger a new scale up after Terraform has scaled down.
I've tried to fix this by setting the node_count to 0 when auto_scaling is enabled, but it didn't work as the node_count needs to be at least 1. Seems like the only way to really "fix" this is to remove the node_count attribute in cluster.tf. A dirty workaround I could imagine would be to have two cluster resources and use count to enable and disable one of them based on the auto_scaling var.
Any inputs on that? Would it be an option to remove the node_count var?