[CI] Optimize nightly CI (#3898)

### What this PR does / why we need it?
This patch mainly fix the the problem of not being able to determine the
exit status of the pod's entrypoint script and some other tiny
optimizations:
1. Shorten wait for server timeout
2. fix typo
3. fix the issue of ais_bench failing to correctly access the proxy URL
in a PD separation scenario.
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.11.0
- vLLM main:
83f478bb19

---------

Signed-off-by: wangli <wangli858794774@gmail.com>

.github/workflows/_e2e_nightly_multi_node.yaml

@@ -60,13 +60,13 @@ defaults:
 # only cancel in-progress runs of the same workflow
 # and ignore the lint / 8 cards test type
 concurrency:
-  group: ascend-nightly-${{ github.ref }}-${{ inputs.config_file_path }}
+  group: ascend-nightly-${{ github.workflow_ref }}-${{ github.ref }}-${{ inputs.config_file_path }}
   cancel-in-progress: true
 
 jobs:
   e2e:
     name: ${{ inputs.config_file_path }}
-    # This is a runner with no NPU for k8s controller
+    # This is the runner with no NPU for k8s controller
     runs-on: ${{ inputs.runner }}
     container:
       image: m.daocloud.io/quay.io/ascend/cann:8.2.rc1-a3-ubuntu22.04-py3.11
@@ -75,7 +75,7 @@ jobs:
         KUBECTL: /root/.cache/.kube/kubectl
         NAMESPACE: vllm-project
         LEADER_POD: vllm-0
-        RESULT_FILE: /root/.cache/tests/ret/test_result.txt
+        RESULT_FILE: /root/.cache/tests/ret_${{ inputs.soc_version }}
     steps:
         - name: Install system denpendencies
           run: |
@@ -84,7 +84,7 @@ jobs:
            pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
            pip install jinja2-cli
 
-           apt-get update -y && apt-get install -y git curl
+           #apt-get update -y && apt-get install -y git curl
 
         - name: Install kubectl
           run: |
@@ -117,8 +117,8 @@ jobs:
           run: |
             # pre clear the crd resources created by lws
             kubectl delete leaderworkerset vllm -n "$NAMESPACE" --ignore-not-found
-
         - name: Launch cluster
+          id: launcher
           run: |
             set -e
 
@@ -130,6 +130,8 @@ jobs:
             vllm_ascend_ref="${{ inputs.vllm_ascend_ref }}"
             vllm_ascend_remote_url="${{ inputs.vllm_ascend_remote_url }}"
             result_file_path="$RESULT_FILE"
+            fail_tag=FAIL_TAG_"${{ inputs.config_file_path }}"
+            echo "FAIL_TAG=${fail_tag}" >> $GITHUB_ENV
 
             required_params=("size" "replicas" "image" "config_file_path")
             for param in "${required_params[@]}"; do
@@ -155,8 +157,7 @@ jobs:
               -D vllm_ascend_ref="$vllm_ascend_ref" \
               -D result_file_path="$result_file_path" \
               -D npu_per_node="$npu_per_node" \
-              -D controller_name="$HOSTNAME" \
-              -D kb_secret=${{ secrets.KUBECONFIG_B64 }} \
+              -D fail_tag="$fail_tag" \
               --outfile lws.yaml
 
             kubectl apply -f ./lws.yaml
@@ -180,7 +181,14 @@ jobs:
 
         - name: Stream logs
           run: |
-            kubectl logs -f "$LEADER_POD" -n "$NAMESPACE"
+            set -euo pipefail
+            echo "Looking for logs containing: $FAIL_TAG"
+            kubectl logs -f "$LEADER_POD" -n "$NAMESPACE" | while read -r line; do
+              echo "$line"
+              if echo "$line" | grep -q "$FAIL_TAG"; then
+                exit 1   # workflow step failed
+              fi
+            done
 
         - name: Post process
           if: always()

.github/workflows/_e2e_nightly_single_node.yaml

@@ -44,7 +44,7 @@ defaults:
 # only cancel in-progress runs of the same workflow
 # and ignore the lint / 1 card / 4 cards test type
 concurrency:
-  group: ascend-nightly-${{ github.ref }}-${{ inputs.tests }}
+  group: ascend-nightly-${{ github.workflow_ref }}-${{ github.ref }}-${{ inputs.tests }}
   cancel-in-progress: true
 
 jobs:

.github/workflows/vllm_ascend_test_nightly_a2.yaml

@@ -42,6 +42,7 @@ concurrency:
 
 jobs:
   single-node-tests:
+    name: single-node
     if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
     strategy:
       fail-fast: false
@@ -63,6 +64,7 @@ jobs:
       tests: ${{ matrix.test_config.tests }}
 
   multi-node-tests:
+    name: multi-node
     needs: single-node-tests
     if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
     strategy:
@@ -71,10 +73,10 @@ jobs:
       matrix:
         test_config:
           - name: multi-node-deepseek-dp
-            config_file_path: tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2.yaml
+            config_file_path: DeepSeek-R1-W8A8-A2.yaml
             size: 2
           - name: multi-node-deepseek-dp-torchair
-            config_file_path: tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2-torchair.yaml
+            config_file_path: DeepSeek-R1-W8A8-A2-torchair.yaml
             size: 2
     uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
     with:

.github/workflows/vllm_ascend_test_nightly_a3.yaml

@@ -42,6 +42,7 @@ concurrency:
 
 jobs:
   single-node-tests:
+    name: single-node
     if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
     strategy:
       fail-fast: false
@@ -85,6 +86,7 @@ jobs:
       tests: ${{ matrix.test_config.tests }}
 
   multi-node-tests:
+    name: multi-node
     needs: single-node-tests
     if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
     strategy:
@@ -93,19 +95,19 @@ jobs:
       matrix:
         test_config:
           - name: multi-node-deepseek-pd
-            config_file_path: tests/e2e/nightly/multi_node/config/models/DeepSeek-V3.yaml
+            config_file_path: DeepSeek-V3.yaml
             size: 2
           - name: multi-node-qwen3-dp
-            config_file_path: tests/e2e/nightly/multi_node/config/models/Qwen3-235B-A3B.yaml
+            config_file_path: Qwen3-235B-A3B.yaml
             size: 2
           - name: multi-node-dpsk-4node-pd
-            config_file_path: tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8.yaml
+            config_file_path: DeepSeek-R1-W8A8.yaml
             size: 4
           - name: multi-node-qwenw8a8-2node
-            config_file_path: tests/e2e/nightly/multi_node/config/models/Qwen3-235B-W8A8.yaml
+            config_file_path: Qwen3-235B-W8A8.yaml
             size: 2
           - name: multi-node-glm-2node
-            config_file_path: tests/e2e/nightly/multi_node/config/models/GLM-4_5.yaml
+            config_file_path: GLM-4_5.yaml
             size: 2
     uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
     with:
@@ -117,12 +119,3 @@ jobs:
       config_file_path: ${{ matrix.test_config.config_file_path }}
     secrets:
       KUBECONFIG_B64: ${{ secrets.KUBECONFIG_B64 }}
-
-  clear_resources:
-    needs: multi-node-tests
-    if: always() && (github.event_name == 'schedule' || github.event_name == 'workflow_dispatch')
-    uses: ./.github/workflows/_kill_lws_resources.yaml
-    with:
-      runner: linux-aarch64-a3-0
-    secrets:
-      KUBECONFIG_B64: ${{ secrets.KUBECONFIG_B64 }}