Use correct node type for GPU test

examples/examples_nodejs_test.go

@@ -878,6 +878,22 @@ func TestAccManagedNodeGroupOS(t *testing.T) {
 
 				assert.NoError(t, utils.ValidateNodePodCapacity(t, info.Outputs["kubeconfig"], 5, 100, "increased-pod-capacity"))
 				assert.NoError(t, utils.ValidateNodeStorage(t, info.Outputs["kubeconfig"], 4, 100*1_000_000_000, "increased-storage-capacity"))
+
+				assert.NoError(t, utils.ValidateNodes(t, info.Outputs["kubeconfig"], func(nodes *corev1.NodeList) {
+					require.NotNil(t, nodes)
+					assert.NotEmpty(t, nodes.Items)
+
+					var foundNodes = 0
+					for _, node := range nodes.Items {
+						if gpus, ok := node.Status.Capacity["nvidia.com/gpu"]; !ok {
+							continue
+						} else {
+							assert.True(t, gpus.CmpInt64(1) == 0)
+							foundNodes++
+						}
+					}
+					assert.Equal(t, 1, foundNodes, "Expected %s nodes with GPU")
+				}))
 			},
 		})
 

examples/tests/managed-ng-os/index.ts

@@ -104,12 +104,12 @@ const managedNodeGroupBottlerocket = eks.createManagedNodeGroup("bottlerocket-mn
   nodeRole: role,
 });
 
-// Create a Bottlerocket node group with GPU support
+// Create a Bottlerocket node group with GPU support (it's the cheapest GPU instance type)
 const managedNodeGroupBottlerocketGpu = eks.createManagedNodeGroup("bottlerocket-mng-gpu", {
   ...scalingConfig,
   cluster: cluster,
   operatingSystem: eks.OperatingSystem.Bottlerocket,
-  instanceTypes: ["t3.medium"],
+  instanceTypes: ["g5g.xlarge"],
   nodeRole: role,
   gpu: true,
 });