Add option to ignore changes to node pool size

[scott4589]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Description

The oci_containerengine_node_pool resource currently has the ignore_changes lifecycle rule applied on various attributes but this does not include node_config_details.size.

This presents an issue when using the cluster autoscaler because it results in Terraform trying to revert the node count in the pool to the value specified by this variable. This behaviour may not be desired if the autoscaler has scaled up/down since the last Terraform apply operation and you want the nodes currently in place to remain unaffected following future apply operations.

The documentation suggests adding the ignore_changes lifecycle rule for the node_config_details.size field, so it would be very useful to have this provided as an option here.

New or Affected Resource(s)

oci_containerengine_node_pool

[devoncrouse]

Hi @scott4589, this is currently intentional, though I agree with the problem generally - it relates to hashicorp/terraform#24188 for ignoring fields dynamically. Without this, the size can never be changed through the module which can be a severe limitation the other way, when the autoscaler may not even be in use.

[shb-mll]

@devoncrouse what is the recommendation in this case - where I have k8s cluster autoscaler enabled and I want to avoid changing the node pool size with subsequent terraform apply?

[12345ieee]

I have the opposite usecase to you that @devoncrouse mentions.

In version 5.x, which has support for the autoscale at the nodepool definition level, this module could create a different nodepool object for autoscaler=true/false with two different ignore lists.

Clearly what we all need is a dynamic terraform lifecycle I hope hashicorp/terraform#24188 gets done soon.