fix: report an error when the node template is not found

Without this change, no instance types are found and the resulting
behavior is a generic error regarding not finding instance types.

pkg/cloudprovider/cloudprovider.go

@@ -177,7 +177,10 @@ func (c *CloudProvider) GetInstanceTypes(ctx context.Context, nodePool *corev1be
 		if errors.IsNotFound(err) {
 			c.recorder.Publish(cloudproviderevents.NodePoolFailedToResolveNodeClass(nodePool))
 		}
-		return nil, client.IgnoreNotFound(fmt.Errorf("resolving node class, %w", err))
+		// We must return an error here in the event of the node class not being found. Otherwise users just get
+		// no instance types and a failure to schedule with no indicator pointing to a bad configuration
+		// as the cause.
+		return nil, fmt.Errorf("resolving node class, %w", err)
 	}
 	// TODO, break this coupling
 	instanceTypes, err := c.instanceTypeProvider.List(ctx, nodePool.Spec.Template.Spec.Kubelet, nodeClass)

pkg/cloudprovider/machine_test.go

@@ -666,12 +666,12 @@ var _ = Describe("Machine/CloudProvider", func() {
 				// select nothing!
 				AMISelector: map[string]string{"Name": "nothing"},
 			})
+			misconfiguredNodeTemplate.Name = "misconfigured"
 			prov2 := test.Provisioner(coretest.ProvisionerOptions{
 				ProviderRef: &v1alpha5.MachineTemplateRef{
 					APIVersion: misconfiguredNodeTemplate.APIVersion,
 					Kind:       misconfiguredNodeTemplate.Kind,
-					// select nothing!
-					Name: "nothing",
+					Name:       "misconfigured",
 				},
 			})
 			ExpectApplied(ctx, env.Client, provisioner, prov2, nodeTemplate, misconfiguredNodeTemplate)