remove code for getting local ip/nic

Signed-off-by: Lizhi Zhou <[email protected]>

Author: reasonsolo

tests/integration/defs/test_e2e.py

@@ -1768,13 +1768,11 @@ def test_trtllm_multimodal_benchmark_serving(llm_root, llm_venv):
 
 
 @pytest.mark.skip_less_device(4)
-@pytest.mark.skip_less_device_memory(80000)
-@pytest.mark.parametrize("ctx_config", ["tp2pp1", "tp1pp2"],
-                         ids=["ctx_tp2pp1", "ctx_tp1pp2"])
-@pytest.mark.parametrize("gen_config", ["tp2pp1", "tp1pp2"],
-                         ids=["gen_tp2pp1", "gen_tp1pp2"])
-def test_openai_disaggregated_serving_multi_nodes(llm_root, llm_venv,
-                                                  ctx_config, gen_config):
+@pytest.mark.skip_less_device_memory(40000)
+@pytest.mark.parametrize("gen_config", ["gen_tp2pp1", "gen_tp1pp2"])
+@pytest.mark.parametrize("ctx_config", ["ctx_tp2pp1", "ctx_tp1pp2"])
+def test_openai_disagg_multi_nodes_completion(llm_root, llm_venv, ctx_config,
+                                              gen_config):
     test_root = unittest_path() / "llmapi" / "apps"
     llm_venv.run_cmd([
         "-m",

tests/integration/test_lists/qa/llm_function_multinode.txt

@@ -7,3 +7,5 @@ test_e2e.py::test_multi_nodes_eval[llama4-models/nvidia/Llama-4-Maverick-17B-128
 test_e2e.py::test_multi_nodes_eval[Qwen3/Qwen3-235B-A22B-tp16-mmlu]
 test_e2e.py::test_multi_nodes_eval[Qwen3/saved_models_Qwen3-235B-A22B_nvfp4_hf-tp16-mmlu]
 test_e2e.py::test_multi_nodes_eval[DeepSeek-R1/DeepSeek-R1-0528-FP4-tp16-mmlu]
+test_e2e.py::test_openai_disagg_multi_nodes_completion[ctx_tp2pp1-gen_tp2pp1]
+test_e2e.py::test_openai_disagg_multi_nodes_completion[ctx_tp1pp2-gen_tp1pp2]

tests/unittest/llmapi/apps/_test_disagg_serving_multi_nodes.py

@@ -1,6 +1,5 @@
 import os
 import socket
-import subprocess
 import time
 
 import openai
@@ -9,7 +8,7 @@
 
 from ..test_llm import get_model_path
 from .openai_server import RemoteDisaggOpenAIServer, RemoteOpenAIServer
-from .utils import expand_slurm_nodelist, get_local_interfaces, get_local_ip
+from .utils import expand_slurm_nodelist
 
 RANK = int(os.environ.get("SLURM_PROCID", 0))
 NODE_RANK = int(os.environ.get("SLURM_NODEID", 0))
@@ -52,56 +51,14 @@ def is_pytest_node():
     return NODE_RANK == 0
 
 
-def find_nic():
-    test_ip = socket.gethostbyname(get_the_other_host())
-    print(f"test_ip: {test_ip} for the other host {get_the_other_host()}")
-    try:
-        # iproute2 may not be installed
-        proc = subprocess.run(["ip", "route", "get", test_ip],
-                              capture_output=True,
-                              text=True,
-                              shell=True,
-                              check=True)
-        output_parts = proc.stdout.split()
-        if "dev" in output_parts:
-            dev_idx = output_parts.index("dev")
-            nic_name = output_parts[dev_idx + 1]
-        else:
-            raise ValueError("Could not find 'dev' in ip route output")
-        print(f"get NIC name from ip route, result: {nic_name}")
-        return nic_name
-    except Exception as e:
-        print(f"Failed to find NIC from ip route: {e}")
-        try:
-            # Establish a socket to the test ip, then get the local ip from the socket,
-            # enumerate the local interfaces and find the one with the local ip
-            local_ip = get_local_ip(test_ip)
-            local_ip_dict = get_local_interfaces()
-            for nic_name, ip in local_ip_dict.items():
-                if ip == local_ip:
-                    return nic_name
-        except OSError as e:
-            print(f"Failed to find NIC from local interfaces: {e}")
-        return None
-
-
 def env():
     # Remove MPI related environment variables to isolate the ctx/gen processes
     # so that they will not be in the same MPI communicator, otherwise the rank and world_size may mismatch
-    new_env = {
+    return {
         k: v
         for k, v in os.environ.items()
         if not ('PMI_' in k or 'OMPI_' in k or 'PMIX_' in k or 'SLURM_' in k)
     }
-    nic = find_nic()
-    if nic:
-        # TODO: integrate this into disagg-serving
-        # setting TRTLLM_UCX_INTERFACE manually if possible because the interfaces found automatically by TRTLLM can have the same ip across nodes, then cache transceiver may fail to send/receive kv cache
-        print(f"setting TRTLLM_UCX_INTERFACE to {nic}")
-        new_env["TRTLLM_UCX_INTERFACE"] = nic
-    else:
-        print(f"Failed to find NIC, will use default UCX interface")
-    return new_env
 
 
 @pytest.fixture(scope="module")

tests/unittest/llmapi/apps/utils.py

@@ -13,11 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import array
-import fcntl
 import re
-import socket
-import struct
 from pathlib import Path
 from typing import Any, Callable
 
@@ -180,45 +176,6 @@ def server_with_custom_sampler(model_name: str, request: Any, backend: str,
     return server_with_custom_sampler
 
 
-def get_local_ip(test_ip):
-    with socket.socket(socket.AF_INET, socket.SOCK_DGRAM) as s:
-        s.connect((test_ip, 80))
-        return s.getsockname()[0]
-
-
-# TODO: Avoid introducing another dependency since this is for test only. If we need to
-# detect UCX interface in the ctx/gen server, we need to support ipv6 and use a package like netifaces
-def get_local_interfaces():
-    """ Returns a dictionary of name:ip key value pairs. """
-    MAX_BYTES = 4096
-    FILL_CHAR = b'\0'
-    SIOCGIFCONF = 0x8912
-    sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
-    names = array.array('B', MAX_BYTES * FILL_CHAR)
-    names_address, names_length = names.buffer_info()
-    mutable_byte_buffer = struct.pack('iL', MAX_BYTES, names_address)
-    mutated_byte_buffer = fcntl.ioctl(sock.fileno(), SIOCGIFCONF,
-                                      mutable_byte_buffer)
-    max_bytes_out, names_address_out = struct.unpack('iL', mutated_byte_buffer)
-    namestr = names.tobytes()
-    namestr[:max_bytes_out]
-    bytes_out = namestr[:max_bytes_out]
-    ip_dict = {}
-    for i in range(0, max_bytes_out, 40):
-        name = namestr[i:i + 16].split(FILL_CHAR, 1)[0]
-        name = name.decode('utf-8')
-        ip_bytes = namestr[i + 20:i + 24]
-        full_addr = []
-        for netaddr in ip_bytes:
-            if isinstance(netaddr, int):
-                full_addr.append(str(netaddr))
-            elif isinstance(netaddr, str):
-                full_addr.append(str(ord(netaddr)))
-        ip_dict[name] = '.'.join(full_addr)
-
-    return ip_dict
-
-
 def expand_slurm_nodelist(nodelist_str):
     """
     Expand SLURM nodelist format into individual node names.