DeepSeek-R1-w8a8多机推理报错ConnectionRefusedError: [Errno 111] Connection refused
收藏回复举报
DeepSeek-R1-w8a8多机推理报错ConnectionRefusedError: [Errno 111] Connection refused
t('forum.solved') 已解决
发表于2025-02-16 00:10:12
0 查看

1. 环境配置

  • 容器:使用Mindie最新镜像mindie:2.0.T3-800I-A2-py311-openeuler24.03-lts-aarch64
    docker run -itd --privileged  --name=mindie --net=host \
     --shm-size 500g \
     --device=/dev/davinci0 \
     --device=/dev/davinci1 \
     --device=/dev/davinci2 \
     --device=/dev/davinci3 \
     --device=/dev/davinci4 \
     --device=/dev/davinci5 \
     --device=/dev/davinci6 \
     --device=/dev/davinci7 \
     --device=/dev/davinci_manager \
     --device=/dev/hisi_hdc \
     --device /dev/devmm_svm \
     --net=atlas-overlay \
     -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
     -v /usr/local/Ascend/firmware:/usr/local/Ascend/firmware \
     -v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
     -v /usr/local/sbin:/usr/local/sbin \
     -v /etc/hccn.conf:/etc/hccn.conf \
     -v /mnt/:/mnt/ \
     25ba5f455ae3 \
     bash

2. 多机通信环境变量设置

export ATB_LLM_HCCL_ENABLE=1
export ATB_LLM_COMM_BACKEND="hccl"
export HCCL_CONNECT_TIMEOUT=7200
export WORLD_SIZE=16
export HCCL_EXEC_TIMEOUT=0
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export MIES_CONTAINER_IP=10.0.0.2
export RANKTABLEFILE=/mnt/zju2/code_aaron/rank_table_file.json

3. rank_table_file.json设置

{
    "version": "1.0",
    "server_count": "2",
    "server_list": [
        {
            "server_id": "172.16.144.144",
            "container_ip": "10.0.0.2",
            "device": [
                { "device_id": "0", "device_ip": "29.10.189.120", "rank_id": "0" }, 
                { "device_id": "1", "device_ip": "29.10.126.84", "rank_id": "1" },
                { "device_id": "2", "device_ip": "29.10.83.192", "rank_id": "2" },
                { "device_id": "3", "device_ip": "29.10.165.249", "rank_id": "3" },
                { "device_id": "4", "device_ip": "29.10.184.45", "rank_id": "4" },
                { "device_id": "5", "device_ip": "29.10.31.246", "rank_id": "5" },
                { "device_id": "6", "device_ip": "29.10.110.72", "rank_id": "6" },
                { "device_id": "7", "device_ip": "29.10.153.77", "rank_id": "7" }
            ]
        },
        {
            "server_id": "172.16.144.161",
            "container_ip": "10.0.0.15",
            "device": [
                { "device_id": "0", "device_ip": "29.10.46.97", "rank_id": "8" },
                { "device_id": "1", "device_ip": "29.10.9.16", "rank_id": "9" },
                { "device_id": "2", "device_ip": "29.10.113.87", "rank_id": "10" },
                { "device_id": "3", "device_ip": "29.10.64.147", "rank_id": "11" },
                { "device_id": "4", "device_ip": "29.10.189.233", "rank_id": "12" },
                { "device_id": "5", "device_ip": "29.10.140.91", "rank_id": "13" },
                { "device_id": "6", "device_ip": "29.10.82.192", "rank_id": "14" },
                { "device_id": "7", "device_ip": "29.10.66.160", "rank_id": "15" }
            ]
        }
    ],
    "status": "completed"
}

4. mindie-server启动

# 以下命令需在所有机器上同时执行
# 解决权重加载过慢问题
export OMP_NUM_THREADS=1
# 设置显存比
export NPU_MEMORY_FRACTION=0.95
# 拉起服务化
cd /usr/local/Ascend/mindie/latest/mindie-service/bin/
./mindieservice_daemon

5. 硬件信息

  • 2台8*910B服务器
  • 每台服务器1.5T内存
  • 使用mindie最新容器部署推理,并且实现两个服务器容器之间ssh免密互连和两台服务器ssh免密互连

6. 报错信息

File "/usr/local/Ascend/ascend-toolkit/latest/python/site-packages/tbe/common/repository_manager/route.py", line 63, in wrapper
    func(*args, **kwargs)
  File "/usr/local/Ascend/ascend-toolkit/latest/python/site-packages/tbe/common/repository_manager/route.py", line 264, in task_distribute
    resource_proxy[SUB_PROCESS_STATE].append(True)
  File "<string>", line 2, in append
  File "/usr/lib64/python3.11/multiprocessing/managers.py", line 819, in _callmethod
    self._connect()
  File "/usr/lib64/python3.11/multiprocessing/managers.py", line 806, in _connect
    conn = self._Client(self._token.address, authkey=self._authkey)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.11/multiprocessing/connection.py", line 518, in Client
    c = SocketClient(address)
        ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.11/multiprocessing/connection.py", line 646, in SocketClient
    s.connect(address)
ConnectionRefusedError: [Errno 111] Connection refused
Process ForkServerProcess-8:
Traceback (most recent call last):
  File "/usr/lib64/python3.11/multiprocessing/managers.py", line 815, in _callmethod
    conn = self._tls.connection
           ^^^^^^^^^^^^^^^^^^^^
AttributeError: 'ForkAwareLocal' object has no attribute 'connection'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/lib64/python3.11/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/usr/lib64/python3.11/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/local/Ascend/ascend-toolkit/latest/python/site-packages/tbe/common/repository_manager/route.py", line 71, in wrapper
    raise exp
  File "/usr/local/Ascend/ascend-toolkit/latest/python/site-packages/tbe/common/repository_manager/route.py", line 63, in wrapper
    func(*args, **kwargs)
  File "/usr/local/Ascend/ascend-toolkit/latest/python/site-packages/tbe/common/repository_manager/route.py", line 264, in task_distribute
    resource_proxy[SUB_PROCESS_STATE].append(True)
  File "<string>", line 2, in append
  File "/usr/lib64/python3.11/multiprocessing/managers.py", line 819, in _callmethod
    self._connect()
  File "/usr/lib64/python3.11/multiprocessing/managers.py", line 806, in _connect
    conn = self._Client(self._token.address, authkey=self._authkey)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.11/multiprocessing/connection.py", line 518, in Client
    c = SocketClient(address)
        ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.11/multiprocessing/connection.py", line 646, in SocketClient
    s.connect(address)
ConnectionRefusedError: [Errno 111] Connection refused

我要发帖子