[CI] Optimize nightly CI (#3858)

### What this PR does / why we need it?
This patch optimize nightly CI:
1. Bug fixes ais_bench get None repo_type error
2. Fix A2 install kubectl error with arm arch
3. Fix the multi_node CI unable to determine whether the job was
successful error
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.11.0rc3
- vLLM main:
83f478bb19

---------

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

tests/e2e/nightly/multi_node/config/multi_node_config.py

@@ -15,7 +15,7 @@ from tests.e2e.nightly.multi_node.config.utils import (get_avaliable_port,
 
 setup_logger()
 logger = logging.getLogger(__name__)
-DISAGGREGATED_PREFILL_PROXY_SCRIPT = "examples/disaggregated_prefill_v1/load_balance_proxy_layerwise_server_example.py"
+DISAGGREGATED_PREFILL_PROXY_SCRIPT = "examples/disaggregated_prefill_v1/load_balance_proxy_server_example.py"
 DISAGGEGATED_PREFILL_PORT = 5333
 
 

tests/e2e/nightly/multi_node/scripts/lws.yaml.jinja2

@@ -7,7 +7,7 @@ spec:
   replicas: {{ replicas | default(1) }}
   leaderWorkerTemplate:
     size: {{ size | default(2) }}
-    restartPolicy: RecreateGroupOnPodRestart
+    restartPolicy: None
     leaderTemplate:
       metadata:
         labels:
@@ -30,6 +30,10 @@ spec:
                 value: {{ vllm_ascend_remote_url | default("https://github.com/vllm-project/vllm-ascend.git") }}
               - name: RESULT_FILE_PATH
                 value: {{ result_file_path | default("/root/.cache/tests/ret/test_result.txt") }}
+              - name: CONTROLLER_NAME
+                value: {{ controller_name | default("placeholder") }}
+              - name: SECRET
+                value: {{ kb_secret | default("placeholder") }}
             command:
               - sh
               - -c

tests/e2e/nightly/multi_node/scripts/run.sh

@@ -162,14 +162,31 @@ kill_npu_processes() {
 }
 
 run_tests() {
-    pytest -sv tests/e2e/nightly/multi_node/test_multi_node.py
+    set +e
     kill_npu_processes
+    pytest -sv tests/e2e/nightly/multi_node/test_multi_node.py
     ret=$?
     if [ "$LWS_WORKER_INDEX" -eq 0 ]; then
-        mkdir -p "$(dirname "$RESULT_FILE_PATH")"
-        echo $ret > "$RESULT_FILE_PATH"
+        if [ $ret -eq 0 ]; then
+            print_success "All tests passed!"
+        else
+            print_error "Some tests failed!"
+            kubectl delete pod $CONTROLLER_NAME -n vllm-project
+        fi
     fi
-    return $ret
+    set -e
+}
+
+install_kubectl() {
+    arch=$(uname -m)
+    KUBECTL=/root/.cache/.kube/kubectl
+    if echo "$arch" | grep -qiE "arm|aarch64"; then
+        echo "Detected ARM architecture: $arch"
+        KUBECTL="$KUBECTL"_arm
+    fi
+    install -o root -g root -m 0755 $KUBECTL /usr/local/bin/kubectl
+    echo "$SECRET" | base64 -d > /tmp/kubeconfig
+    export KUBECONFIG=/tmp/kubeconfig
 }
 
 main() {
@@ -177,6 +194,7 @@ main() {
     check_and_config
     checkout_src
     install_sys_dependencies
+    install_kubectl
     install_vllm
     install_ais_bench
     # to speed up mooncake build process, install Go here