"Unset" value for google_compute_region_instance_group_manager target_size argument?

[danisla]

Related issues: #65, #13. Terraform #5471

Affected resources:

• google_compute_region_instance_group_manager
• google_compute_instance_group_manager

The target_size argument is mutually exclusive from the autoscaler resource. Is there a way to conditionally set the target_size argument so that when I'm using a autoscaler, it can be unset?

Setting it to0 forces my instance group to 0, which conflicts with the autoscaler.

target_size = "${var.autoscaling ? 0 : var.size}"

If 0 was the unset value, then the above indented use case would work. If 0 is valid, can we use -1?

[danawillow]

Taking a look at what the cloud console claims a REST request for MIG with autoscaling is, I think the "unset" value is actually 1. Can you give that a try @danisla and see if that works?

And now it makes sense to me why 1 was originally the default here. I think the right answer for the future is to get rid of the Computed on it and instead set a default of 1, but leave in the ForceSendFields so people can set zero that want to. @rileykarson, if you have a moment, do you remember off the top of your head whether that was something you had tried? I just gave it a quick go and it looks like tests pass just fine.

[danisla]

@danawillow, I set the target_size = 1 and autoscaling min_replicas = 3 and terraform plan wants to change the target_size from 3 back to 1

Terraform will perform the following actions:

  ~ module.mig1.google_compute_region_instance_group_manager.default
      target_size: "3" => "1"

I get an error if I try to apply it:

Error: Error applying plan:

1 error(s) occurred:

* module.mig1.google_compute_region_instance_group_manager.default: 1 error(s) occurred:

* google_compute_region_instance_group_manager.default: Error resizing RegionInstanceGroupManager: googleapi: Error 400: The instance_group_manager resource 'projects/disla-terraform-admin-101/regions/us-central1/instanceGroupManagers/autoscale-cluster' is already being used by 'projects/disla-terraform-admin-101/regions/us-central1/autoscalers/autoscale-cluster', resourceInUseByAnotherResource

I have a hack workaround now, which forces the target_size to follow the min_replicas, I think this will only work if the cluster is at it's minimum scale though.

target_size = "${var.autoscaling ? var.min_replicas : var.size}"

[danawillow]

Ah ok, that makes sense, and now I see why this was so tricky. We want target_size to be computed so that if it's unset it can just go ahead and change. But we also want to be able to specify a size of zero. Unfortunately, terraform can't currently tell the difference between unset and set to zero, so I don't think there's anything I can do in the google provider at the moment to fix this.

@apparentlymart you might be interested in this since I know you're working on improvements in this area.

[danawillow]

I haven't actually checked, but this might be fixable now either with GetOkExists or CustomDiff. I'm not working on it right now though so unassigning.

[paddycarver]

This should be resolved by the upcoming 0.12 changes.

[rileykarson]

This should be possible in 0.12 with null.

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you feel this issue should be reopened, we encourage creating a new issue linking back to this one for added context. If you feel I made an error 🤖 🙉 , please reach out to my human friends 👉 hashibot-feedback@hashicorp.com. Thanks!