From 1e08ee74e577f4988d0c06ece4310bbff7e2050e Mon Sep 17 00:00:00 2001
From: YuBaoku <49938469+EmmonsCurse@users.noreply.github.com>
Date: Mon, 13 Apr 2026 20:23:49 +0800
Subject: [PATCH] [CI] Modify 4-card container startup config and move test
 case (#7363)

---
 .github/workflows/_gpu_4cards_case_test.yml          | 12 +++++++++++-
 .../test_ernie_03b_pd_router_v1_rdma_tp2.py          |  6 +++---
 2 files changed, 14 insertions(+), 4 deletions(-)
 rename tests/e2e/{ => 4cards_cases}/test_ernie_03b_pd_router_v1_rdma_tp2.py (98%)

diff --git a/.github/workflows/_gpu_4cards_case_test.yml b/.github/workflows/_gpu_4cards_case_test.yml
index 5c9a51aa80..797fddb775 100644
--- a/.github/workflows/_gpu_4cards_case_test.yml
+++ b/.github/workflows/_gpu_4cards_case_test.yml
@@ -181,11 +181,17 @@ jobs:
             docker rm -f ${runner_name} || true
           fi
 
+          export RDMA_DEVICES=$(find /dev/infiniband/uverbs* -maxdepth 1 -not -type d | xargs -I{} echo '--device {}:{}')
+
           docker run --rm --net=host \
-          --shm-size=64g \
           --sysctl kernel.msgmax=1048576 \
           --sysctl kernel.msgmnb=268435456 \
           --name ${runner_name} \
+          --cap-add=SYS_PTRACE --cap-add=IPC_LOCK \
+          --shm-size=64G \
+          ${RDMA_DEVICES} \
+          --device=/dev/infiniband/rdma_cm \
+          --ulimit memlock=-1:-1 \
           -v $(pwd):/workspace -w /workspace \
           -v "${CACHE_DIR}/gitconfig:/etc/gitconfig:ro" \
           -v "${CACHE_DIR}/.cache:/root/.cache" \
@@ -197,6 +203,10 @@ jobs:
           -e "FD_METRICS_PORT=${FD_METRICS_PORT}" \
           -e "FLASK_PORT=${FLASK_PORT}" \
           -e "FD_CACHE_QUEUE_PORT=${FD_CACHE_QUEUE_PORT}" \
+          -e "FD_ROUTER_PORT=${FD_ROUTER_PORT}" \
+          -e "FD_CONNECTOR_PORT=${FD_CONNECTOR_PORT}" \
+          -e "FD_RDMA_PORT=${FD_RDMA_PORT}" \
+          -e "CLEAN_CUDA=1" \
           -e TZ="Asia/Shanghai" \
           -e "fd_wheel_url=${fd_wheel_url}" \
           -e "BASE_REF=${BASE_REF}" \
diff --git a/tests/e2e/test_ernie_03b_pd_router_v1_rdma_tp2.py b/tests/e2e/4cards_cases/test_ernie_03b_pd_router_v1_rdma_tp2.py
similarity index 98%
rename from tests/e2e/test_ernie_03b_pd_router_v1_rdma_tp2.py
rename to tests/e2e/4cards_cases/test_ernie_03b_pd_router_v1_rdma_tp2.py
index 0bbc8186a5..85858da1dc 100644
--- a/tests/e2e/test_ernie_03b_pd_router_v1_rdma_tp2.py
+++ b/tests/e2e/4cards_cases/test_ernie_03b_pd_router_v1_rdma_tp2.py
@@ -26,7 +26,7 @@ import time
 
 import pytest
 import requests
-from utils.serving_utils import (
+from e2e.utils.serving_utils import (
     FD_API_PORT,
     FD_CACHE_QUEUE_PORT,
     FD_ENGINE_QUEUE_PORT,
@@ -90,7 +90,7 @@ def setup_and_run_server():
 
     # get rdma nics
     current_dir = os.path.dirname(os.path.abspath(__file__))
-    shell_path = os.path.join(current_dir, "utils/get_rdma_nics.sh")
+    shell_path = os.path.join(current_dir, "../utils/get_rdma_nics.sh")
     output = subprocess.check_output(["bash", shell_path, "gpu"], text=True)
     _, rdma_nics = output.split("=")
     print(f"shell_path: {shell_path}, rdma_nics: {rdma_nics}")
@@ -171,7 +171,7 @@ def setup_and_run_server():
     # decode实例
     print("start decode...")
     env_decode = os.environ.copy()
-    env_decode["CUDA_VISIBLE_DEVICES"] = "1"
+    env_decode["CUDA_VISIBLE_DEVICES"] = "2"
     env_decode["FD_LOG_DIR"] = os.path.join(base_log_dir, "log_decode")
     # env_decode["KVCACHE_RDMA_NICS"] = rdma_nics