fix: Set the cluster status as in_progress at the end of update_nodeg…

…roup handler

nodegroup_conductor.Handler.nodegroup_update() sets the cluster obj status
as `UPDATE_IN_PROGRESS` and saves it immediately.
But it only sets the nodegroup(ng) obj status as `UPDATE_IN_PROGRESS` and
saves it after cluster_driver.update_nodegroup() finished.
On the other side, Magnum periodic ClusterUpdateJob will trigger
cluster_driver.update_cluster_status().

In the driver, if update_cluster_status() has been finished before
update_nodegroup(), the nodegroup status could be set as `UPDATE_IN_PROGRESS`
after cluster status has been set to `UPDATE_COMPLETED`.
It means update_cluster_status() is never called again which is
responsible for ng status update even if ng status is not COMPLETED.

This PR saves the ng status first in the update_nodegroup handler, and
sets the cluster status as in_progress again at the end to requeue the
cluster in the periodic ClusterUpdateJob again.

magnum_cluster_api/driver.py

@@ -287,7 +287,15 @@ def update_nodegroup_status(self, context, cluster, nodegroup):
     def update_nodegroup(self, context, cluster, nodegroup):
         # TODO
 
+        # NOTE(okozachenko1203): First we save the nodegroup status because update_cluster_status()
+        #                        could be finished before update_nodegroup().
+        nodegroup.save()
         resources.apply_cluster_from_magnum_cluster(context, self.k8s_api, cluster)
+        # NOTE(okozachenko1203): We set the cluster status as UPDATE_IN_PROGRESS again at the end because
+        #                        update_cluster_status() could be finished and cluster status has been set as
+        #                        UPDATE_COMPLETE before nodegroup_conductor.Handler.nodegroup_update finished.
+        cluster.status = "UPDATE_IN_PROGRESS"
+        cluster.save()
 
     def delete_nodegroup(self, context, cluster, nodegroup):
         nodegroup.status = "DELETE_IN_PROGRESS"