Omit capacity when creating basic plans

The new [Basic SKU][0] does not support any scaling, and it is an error to
provide `Capacity` or `AutoScaleSettings`. The resource is implemented
in such a way that it was not possible to call
`DeploymentsCreateOrUpdateThenPoll` without one of these in the
`ScalingProperties`, which is illegal for the Basic SKUs.

Currently the backend has a hack to silently drop `ScalingProperties`
for Basic SKUs, but this is a confusing user experience:

- initial `terraform apply` submits a `PUT` with `"scalingProperties":
{"capacity": 20}` - this validation error is currently suppressed and
the server reports back `"scalingProperties": null`

- next `terraform plan` detects a `capacity 0 -> 20`

- applying that plan requests a `PATCH` with `"scalingProperties":
{"capacity": 20}`, that fails with a `UnsupportedOnBasicPlan` error

It gets more misleading if users specify a capacity explicity;
`capacity = 500` or `auto_scale_profile` will apply successfully, but
get silently ignored server-side.

This patch checks for basic plans and omits the `scalingProperties`
when the user hasn't changed from the default value. We already
instruct users to ignore `capacity`, but the check is needed to
prevent the implicit default value of 20 from getting into the
`DeploymentsCreateOrUpdateThenPoll` request.

[0]: https://docs.nginx.com/nginxaas/azure/billing/overview/#basic-plan

internal/services/nginx/nginx_deployment_resource.go

@@ -6,6 +6,7 @@ package nginx
 import (
 	"context"
 	"fmt"
+	"strings"
 	"time"
 
 	"github.com/hashicorp/go-azure-helpers/lang/pointer"
@@ -20,6 +21,8 @@ import (
 	"github.com/hashicorp/terraform-provider-azurerm/internal/tf/validation"
 )
 
+const defaultCapacity = 20
+
 type FrontendPrivate struct {
 	IpAddress        string `tfschema:"ip_address"`
 	AllocationMethod string `tfschema:"allocation_method"`
@@ -82,6 +85,9 @@ type DeploymentModel struct {
 	Tags          map[string]string `tfschema:"tags"`
 }
 
+func (m *DeploymentModel) isBasic() bool            { return strings.HasPrefix(m.Sku, "basic") }
+func (m *DeploymentModel) hasDefaultCapacity() bool { return m.Capacity == defaultCapacity }
+
 type DeploymentResource struct{}
 
 var _ sdk.ResourceWithUpdate = (*DeploymentResource)(nil)
@@ -123,7 +129,7 @@ func (m DeploymentResource) Arguments() map[string]*pluginsdk.Schema {
 			Type:          pluginsdk.TypeInt,
 			Optional:      true,
 			ConflictsWith: []string{"auto_scale_profile"},
-			Default:       20,
+			Default:       defaultCapacity,
 			ValidateFunc:  validation.IntPositive,
 		},
 
@@ -430,7 +436,10 @@ func (m DeploymentResource) Create() sdk.ResourceFunc {
 				prop.NetworkProfile.NetworkInterfaceConfiguration.SubnetId = pointer.FromString(model.NetworkInterface[0].SubnetId)
 			}
 
-			if model.Capacity > 0 {
+			// scaling is not supported for basic plans. Ignore default capacity
+			// for basic plans requests, which happen if a user does not
+			// configure capacity in their tf
+			if model.Capacity > 0 && !(model.hasDefaultCapacity() && model.isBasic()) {
 				prop.ScalingProperties = &nginxdeployment.NginxDeploymentScalingProperties{
 					Capacity: pointer.FromInt64(model.Capacity),
 				}

website/docs/r/nginx_deployment.html.markdown

@@ -88,7 +88,7 @@ The following arguments are supported:
 
 * `sku` - (Required) Specifies the NGINX Deployment SKU. Possible values are `standard_Monthly` and `basic_Monthly`. Changing this forces a new resource to be created.
 
--> **NOTE:** If you are setting the `sku` to `basic_Monthly`, you should use [Terraform's `ignore_changes` functionality](https://www.terraform.io/language/meta-arguments/lifecycle#ignore_changes) to ignore changes to the `capacity` field.
+-> **NOTE:** If you are setting the `sku` to `basic_Monthly`, setting a non-default `capacity` or `auto_scale_profile` will fail. You should use [Terraform's `ignore_changes` functionality](https://www.terraform.io/language/meta-arguments/lifecycle#ignore_changes) to ignore changes to the `capacity` field.
 
 * `managed_resource_group` - (Optional) Specify the managed resource group to deploy VNet injection related network resources. Changing this forces a new NGINX Deployment to be created.