Cause node pool to be written to statefile even if creation fails. (#…

…3476) (#2038)

Signed-off-by: Modular Magician <[email protected]>

.changelog/3476.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+container: Fix occasional error with `container_node_pool` partially-successful creations not being recorded if an error occurs on the GCP side.
+```

google-beta/resource_container_node_pool.go

@@ -267,6 +267,11 @@ func resourceContainerNodePoolCreate(d *schema.ResourceData, meta interface{}) e
 	timeout := d.Timeout(schema.TimeoutCreate)
 	startTime := time.Now()
 
+	// Set the ID before we attempt to create - that way, if we receive an error but
+	// the resource is created anyway, it will be refreshed on the next call to
+	// apply.
+	d.SetId(fmt.Sprintf("projects/%s/locations/%s/clusters/%s/nodePools/%s", nodePoolInfo.project, nodePoolInfo.location, nodePoolInfo.cluster, nodePool.Name))
+
 	var operation *containerBeta.Operation
 	err = resource.Retry(timeout, func() *resource.RetryError {
 		operation, err = config.clientContainerBeta.
@@ -287,8 +292,6 @@ func resourceContainerNodePoolCreate(d *schema.ResourceData, meta interface{}) e
 	}
 	timeout -= time.Since(startTime)
 
-	d.SetId(fmt.Sprintf("projects/%s/locations/%s/clusters/%s/nodePools/%s", nodePoolInfo.project, nodePoolInfo.location, nodePoolInfo.cluster, nodePool.Name))
-
 	waitErr := containerOperationWait(config,
 		operation, nodePoolInfo.project,
 		nodePoolInfo.location, "creating GKE NodePool", timeout)