From cd8e5be7c7bb1860e00e6a41cad8bb1db6ce133a Mon Sep 17 00:00:00 2001
From: Li Wang <wangli858794774@gmail.com>
Date: Thu, 4 Dec 2025 23:51:16 +0800
Subject: [PATCH] [Bugfix] Quick hot fix for nightly CI (#4727)

Quick fix for multi-node tests

---------

Signed-off-by: wangli <wangli858794774@gmail.com>
---
 .../vllm_ascend_test_nightly_a2.yaml          | 53 ++++++++++---------
 .../e2e/nightly/multi_node/test_multi_node.py |  6 +--
 2 files changed, 30 insertions(+), 29 deletions(-)

diff --git a/.github/workflows/vllm_ascend_test_nightly_a2.yaml b/.github/workflows/vllm_ascend_test_nightly_a2.yaml
index f098ccc0..01ee56ca 100644
--- a/.github/workflows/vllm_ascend_test_nightly_a2.yaml
+++ b/.github/workflows/vllm_ascend_test_nightly_a2.yaml
@@ -42,35 +42,9 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  multi-node-tests:
-    name: multi-node
-    if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
-    strategy:
-      fail-fast: false
-      max-parallel: 1
-      matrix:
-        test_config:
-          - name: multi-node-deepseek-dp
-            config_file_path: DeepSeek-R1-W8A8-A2.yaml
-            size: 2
-          - name: multi-node-deepseek-dp-torchair
-            config_file_path: DeepSeek-R1-W8A8-A2-torchair.yaml
-            size: 2
-    uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
-    with:
-      soc_version: a2
-      runner: linux-aarch64-a2-0
-      image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a2'
-      replicas: 1
-      size: ${{ matrix.test_config.size }}
-      config_file_path: ${{ matrix.test_config.config_file_path }}
-    secrets:
-      KUBECONFIG_B64: ${{ secrets.KUBECONFIG_A2_B64 }}
-
   single-node-tests:
     name: single-node
     if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
-    needs: multi-node-tests
     strategy:
       fail-fast: false
       matrix:
@@ -139,3 +113,30 @@ jobs:
       model_list: ${{ toJson(matrix.test_config.model_list) }}
       image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:8.3.rc2-910b-ubuntu22.04-py3.11'
       upload: false
+
+
+  multi-node-tests:
+    name: multi-node
+    if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
+    needs: single-node-tests
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        test_config:
+          - name: multi-node-deepseek-dp
+            config_file_path: DeepSeek-R1-W8A8-A2.yaml
+            size: 2
+          - name: multi-node-deepseek-dp-torchair
+            config_file_path: DeepSeek-R1-W8A8-A2-torchair.yaml
+            size: 2
+    uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
+    with:
+      soc_version: a2
+      runner: linux-aarch64-a2-0
+      image: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend:nightly-a2'
+      replicas: 1
+      size: ${{ matrix.test_config.size }}
+      config_file_path: ${{ matrix.test_config.config_file_path }}
+    secrets:
+      KUBECONFIG_B64: ${{ secrets.KUBECONFIG_A2_B64 }}
diff --git a/tests/e2e/nightly/multi_node/test_multi_node.py b/tests/e2e/nightly/multi_node/test_multi_node.py
index 212ad26d..66745a97 100644
--- a/tests/e2e/nightly/multi_node/test_multi_node.py
+++ b/tests/e2e/nightly/multi_node/test_multi_node.py
@@ -100,7 +100,7 @@ async def test_multi_node() -> None:
     disaggregated_prefill = config.disaggregated_prefill
     server_port = config.server_port
     proxy_port = config.proxy_port
-    server_host = config.node_info.ip
+    server_host = config.master_ip
     proxy_script = config.envs.get("DISAGGREGATED_PREFILL_PROXY_SCRIPT", \
         'examples/disaggregated_prefill_v1/load_balance_proxy_server_example.py')
     with config.launch_server_proxy(proxy_script):
@@ -123,8 +123,8 @@ async def test_multi_node() -> None:
                 run_aisbench_cases(local_model_path,
                                    port,
                                    aisbench_cases,
-                                   host_ip=config.cluster_ips[0])
+                                   host_ip=config.master_ip)
             else:
                 # for the nodes except master, should hang until the task complete
-                master_url = f"http://{config.cluster_ips[0]}:{server_port}/health"
+                master_url = f"http://{config.master_ip}:{server_port}/health"
                 remote_server.hang_until_terminated(master_url)