Confused: agent_nodepools and autoscaler_nodepools

[tgrushka]

Thanks for this project!

Not clear from docs: Are agent_nodepools and autoscaler_nodepools complementary or distinct? Are they both worker nodepools, where one has CA autoscaling enabled and the other does not? Can I leave agent_nodepools empty and put all my nodepools in autoscaler_nodepools if I want to be able to scale all of them, and just set min_nodes = 0 and max_nodes = 0 to not use for now, and min_nodes = 1 and max_nodes = 1 to not autoscale a particular pool?

Or, do I need to define all the nodepools in agent_nodepools and then duplicate them in autoscaler_nodepools?

Please also clarify this in the README and example.tf.

Also, maybe note that agent nodes == worker nodes as someone at some point decided to muddy the waters by introducing both terms.

Another related question:
Why not just have agent_nodepools combined, and set autoscaling to true or false on each one, with min/max / count ... or better yet just forget count and have min/max and CA enabled on all by default (maybe this is a resource question)?

[pat-s]

I was also wondering about this.
While agent_nodepools cannot be undefined, one can set count=0 in the first node pool definition and by that effectively avoid deploying non-autoscaled nodepools.

I don't see any downside in only deploying autoscaled pools with a sensible min_size.

This point deserves some/more discussion in the current template.