added test and fix

ITISFoundation · Aug 23, 2024 · 161481c · 161481c
1 parent 40d1ecb
commit 161481c
Show file tree

Hide file tree

Showing 2 changed files with 28 additions and 3 deletions.
diff --git a/services/autoscaling/src/simcore_service_autoscaling/modules/auto_scaling_core.py b/services/autoscaling/src/simcore_service_autoscaling/modules/auto_scaling_core.py
@@ -200,14 +200,13 @@ async def _make_pending_buffer_ec2s_join_cluster(
     app: FastAPI,
     cluster: Cluster,
 ) -> Cluster:
-    # started buffer instance shall be asked to join the cluster once they are running
-    ssm_client = get_ssm_client(app)
-
     if buffer_ec2s_pending := [
         i.ec2_instance
         for i in cluster.pending_ec2s
         if is_buffer_machine(i.ec2_instance.tags)
     ]:
+        # started buffer instance shall be asked to join the cluster once they are running
+        ssm_client = get_ssm_client(app)
         buffer_ec2_connection_state = await limited_gather(
             *[
                 ssm_client.is_instance_connected_to_ssm_server(i.id)

diff --git a/services/autoscaling/tests/unit/test_modules_auto_scaling_computational.py b/services/autoscaling/tests/unit/test_modules_auto_scaling_computational.py
@@ -202,6 +202,32 @@ async def test_cluster_scaling_with_no_tasks_does_nothing(
     )
 
 
+@pytest.mark.acceptance_test(
+    "Ensure this does not happen https://github.com/ITISFoundation/osparc-simcore/issues/6227"
+)
+async def test_cluster_scaling_with_disabled_ssm_does_not_block_autoscaling(
+    minimal_configuration: None,
+    disabled_ssm: None,
+    app_settings: ApplicationSettings,
+    initialized_app: FastAPI,
+    mock_launch_instances: mock.Mock,
+    mock_terminate_instances: mock.Mock,
+    mock_rabbitmq_post_message: mock.Mock,
+    dask_spec_local_cluster: distributed.SpecCluster,
+):
+    await auto_scale_cluster(
+        app=initialized_app, auto_scaling_mode=ComputationalAutoscaling()
+    )
+    mock_launch_instances.assert_not_called()
+    mock_terminate_instances.assert_not_called()
+    _assert_rabbit_autoscaling_message_sent(
+        mock_rabbitmq_post_message,
+        app_settings,
+        initialized_app,
+        dask_spec_local_cluster.scheduler_address,
+    )
+
+
 async def test_cluster_scaling_with_task_with_too_much_resources_starts_nothing(
     minimal_configuration: None,
     app_settings: ApplicationSettings,