From 2708b3abbced02b37d531ff52c1a52f4f4c16423 Mon Sep 17 00:00:00 2001
From: Max Fitton <maxfitton@anyscale.com>
Date: Mon, 30 Nov 2020 18:43:09 -0800
Subject: [PATCH] [Dashboard][Bug] Fix duplicate node total rows in dashboard
 (#12410)

* Fix duplicate node total rows in dashboard by changing the react key of the NodeTotalRow component from the node IP to the node ID (node IP can be duplicated in the case of docker).

* simplify a piece of test code and fix a flaky time out

* lint
---
 .../pages/dashboard/node-info/NodeInfo.tsx    |  2 +-
 dashboard/tests/test_dashboard.py             | 36 ++++++++-----------
 2 files changed, 15 insertions(+), 23 deletions(-)

diff --git a/dashboard/client/src/pages/dashboard/node-info/NodeInfo.tsx b/dashboard/client/src/pages/dashboard/node-info/NodeInfo.tsx
index be9d3787e..2b1a12c03 100644
--- a/dashboard/client/src/pages/dashboard/node-info/NodeInfo.tsx
+++ b/dashboard/client/src/pages/dashboard/node-info/NodeInfo.tsx
@@ -88,7 +88,7 @@ const makeGroupedTableContents = (
     );
     return (
       <NodeRowGroup
-        key={node.ip}
+        key={node.raylet.nodeId}
         node={node}
         workerFeatureData={sortedClusterWorkers}
         features={nodeInfoFeatures}
diff --git a/dashboard/tests/test_dashboard.py b/dashboard/tests/test_dashboard.py
index faa21795d..a37fd3012 100644
--- a/dashboard/tests/test_dashboard.py
+++ b/dashboard/tests/test_dashboard.py
@@ -16,12 +16,9 @@ import redis
 import requests
 
 from ray import ray_constants
-from ray.test_utils import (
-    format_web_url,
-    wait_for_condition,
-    wait_until_server_available,
-    run_string_as_driver,
-)
+from ray.test_utils import (format_web_url, wait_for_condition,
+                            wait_until_server_available, run_string_as_driver,
+                            wait_until_succeeded_without_exception)
 from ray.autoscaler._private.util import (DEBUG_AUTOSCALING_STATUS,
                                           DEBUG_AUTOSCALING_ERROR)
 import ray.new_dashboard.consts as dashboard_consts
@@ -453,22 +450,17 @@ def test_get_cluster_status(ray_start_with_dashboard):
 
     # Check that the cluster_status endpoint works without the underlying data
     # from the GCS, but returns nothing.
-    last_exception = None
-    for _ in range(5):
-        try:
-            response = requests.get(f"{webui_url}/api/cluster_status")
-            response.raise_for_status()
-            assert response.json()["result"]
-            assert "autoscalingStatus" in response.json()["data"]
-            assert response.json()["data"]["autoscalingStatus"] is None
-            assert "autoscalingError" in response.json()["data"]
-            assert response.json()["data"]["autoscalingError"] is None
-            break
-        except requests.RequestException as e:
-            last_exception = e
-            time.sleep(1)
-    else:
-        raise last_exception
+    def get_cluster_status():
+        response = requests.get(f"{webui_url}/api/cluster_status")
+        response.raise_for_status()
+        assert response.json()["result"]
+        assert "autoscalingStatus" in response.json()["data"]
+        assert response.json()["data"]["autoscalingStatus"] is None
+        assert "autoscalingError" in response.json()["data"]
+        assert response.json()["data"]["autoscalingError"] is None
+
+    wait_until_succeeded_without_exception(get_cluster_status,
+                                           (requests.RequestException, ))
 
     # Populate the GCS field, check that the data is returned from the
     # endpoint.