fix: deploy autoscaler right before the cluster creation completed (#307

)

magnum_cluster_api/driver.py

@@ -60,7 +60,9 @@ def create_cluster(self, context, cluster, cluster_create_timeout):
             context, self.k8s_api, cluster
         ).apply()
 
-        resources.apply_cluster_from_magnum_cluster(context, self.k8s_api, cluster)
+        resources.apply_cluster_from_magnum_cluster(
+            context, self.k8s_api, cluster, skip_auto_scaling_release=True
+        )
 
     def _get_cluster_status_reason(self, capi_cluster):
         capi_cluster_status_reason = ""
@@ -127,6 +129,14 @@ def update_cluster_status(self, context, cluster, use_admin_ctx=False):
             )
             cluster.coe_version = capi_cluster.obj["spec"]["topology"]["version"]
 
+            # NOTE(oleks): To avoid autoscaler crashes, we deploy it after the
+            #              cluster api endpoint is reachable.
+            if (
+                cluster.status == "CREATE_IN_PROGRESS"
+                and utils.get_auto_scaling_enabled(cluster)
+            ):
+                resources.ClusterAutoscalerHelmRelease(self.k8s_api, cluster).apply()
+
             for ng in node_groups:
                 if not ng.status.endswith("_COMPLETE"):
                     return