By

总阅读次

文章目录
  1. 1. CreateSession still waiting for response from worker: /job:worker/replica:0/task:0
  2. 2. ImportError: /lib64/libstdc++.so.6: version `CXXABI_1.3.9’ not found
  3. 3. 分布式Tensorflow, ps端运行出现tensorflow.python.framework.errors_impl.UnavailableError: OS Error
  • 参考文献

    • 本文记录笔者在Tensorflow使用上的一些错误的集锦，方便后来人迅速查阅解决问题。

    我是留白。

    我是留白。

    CreateSession still waiting for response from worker: /job:worker/replica:0/task:0

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    2018-12-05 22:18:24.565303: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job ps -> {0 -> localhost:3376}
    2018-12-05 22:18:24.565372: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job worker -> {0 -> localhost:3330, 1 -> localhost:3331}
    2018-12-05 22:18:24.569212: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:381] Started server with target: grpc://localhost:3376
    2018-12-05 22:18:26.170901: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job ps -> {0 -> localhost:3376}
    2018-12-05 22:18:26.170969: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job worker -> {0 -> localhost:3330, 1 -> localhost:3331}
    2018-12-05 22:18:26.174856: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:381] Started server with target: grpc://localhost:3330
    2018-12-05 22:18:27.177003: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job ps -> {0 -> localhost:3376}
    2018-12-05 22:18:27.177071: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job worker -> {0 -> localhost:3330, 1 -> localhost:3331}
    2018-12-05 22:18:27.180980: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:381] Started server with target: grpc://localhost:3331
    2018-12-05 22:18:34.625459: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:worker/replica:0/task:0
    2018-12-05 22:18:34.625513: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:worker/replica:0/task:1
    2018-12-05 22:18:36.231936: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:ps/replica:0/task:0
    2018-12-05 22:18:36.231971: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:worker/replica:0/task:1
    2018-12-05 22:18:37.235899: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:ps/replica:0/task:0
    2018-12-05 22:18:37.235952: I tensorflow/core/distributed_runtime/master.cc:267] CreateSession still waiting for response from worker: /job:worker/replica:0/task:0

    首先保证job_name,task_index,ps_hosts,worker_hosts这四个参数都是正确的,考虑以下这种情况是不正确的：
    在一个IP为192.168.1.100的机器上启动ps或worker进程：

    1
    2
    3
    4
    --job_name=worker
    --task_index=1
    --ps_hosts=192.168.1.100:2222,192.168.1.101:2222
    --worker_hosts=192.168.1.100:2223,192.168.1.101:2223

    因为该进程启动位置是192.168.1.100，但是运行参数中指定的task_index为1，对应的IP地址是ps_hosts或worker_hosts的第二项（第一项的task_index为0)，也就是192.168.1.101，和进程本身所在机器的IP不一致。

    另外一种情况也会导致该问题的发生，从TensorFlow-1.4开始，分布式会自动使用环境变量中的代理去连接，如果运行的节点之间不需要代理互连，那么将代理的环境变量移除即可，在脚本的开始位置添加代码：
    注意这段代码必须写在import tensorflow as tf或者import moxing.tensorflow as mox之前

    1
    2
    3
    import os
    os.enrivon.pop('http_proxy')
    os.enrivon.pop('https_proxy')

    — 摘自(https://bbs.huaweicloud.com/blogs/463145f7a1d111e89fc57ca23e93a89f)

    ImportError: /lib64/libstdc++.so.6: version `CXXABI_1.3.9’ not found

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    /home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
      from ._conv import register_converters as _register_converters
    Traceback (most recent call last):
      File "trainer.py", line 14, in <module>
        import sklearn.datasets
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/sklearn/__init__.py", line 134, in <module>
        from .base import clone
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/sklearn/base.py", line 11, in <module>
        from scipy import sparse
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/scipy/sparse/__init__.py", line 229, in <module>
        from .csr import *
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/scipy/sparse/csr.py", line 15, in <module>
        from ._sparsetools import csr_tocsc, csr_tobsr, csr_count_blocks, \
    ImportError: /lib64/libstdc++.so.6: version `CXXABI_1.3.9' not found (required by /home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/scipy/sparse/_sparsetools.cpython-36m-x86_64-linux-gnu.so)

    系统的库文件较老，不含CXXABI_1.3.9，可将<Anaconda_PATH>/lib加入LD_LIBRARY_PATH中，像这样：

    1
    export LD_LIBRARY_PATH=/home/../anaconda3/lib:$LD_LIBRARY_PATH

    如此，系统会先找到anaconda里面的lib，从而满足要求。

    参考：Stackoverflow. 问题2

    分布式Tensorflow, ps端运行出现tensorflow.python.framework.errors_impl.UnavailableError: OS Error

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    30
    31
    32
    33
    34
    35
    2018-12-07 15:40:05.167922: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job ps -> {0 -> localhost:3333}
    2018-12-07 15:40:05.167970: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:222] Initialize GrpcChannelCache for job worker -> {0 -> 192.168.100.36:3333, 1 -> 192.168.100.37:3333}
    2018-12-07 15:40:05.171857: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:381] Started server with target: grpc://localhost:3333
    Parameter server: waiting for cluster connection...
    2018-12-07 15:40:05.213496: E tensorflow/core/distributed_runtime/master.cc:315] CreateSession failed because worker /job:worker/replica:0/task:0 returned error: Unavailable: OS Error
    2018-12-07 15:40:05.213645: E tensorflow/core/distributed_runtime/master.cc:315] CreateSession failed because worker /job:worker/replica:0/task:1 returned error: Unavailable: OS Error
    Traceback (most recent call last):
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1334, in _do_call
        return fn(*args)
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1317, in _run_fn
        self._extend_graph()
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1352, in _extend_graph
        tf_session.ExtendSession(self._session)
    tensorflow.python.framework.errors_impl.UnavailableError: OS Error

    During handling of the above exception, another exception occurred:

    Traceback (most recent call last):
      File "trainer.py", line 364, in <module>
        tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 125, in run
        _sys.exit(main(argv))
      File "trainer.py", line 70, in main
        num_classes=num_classes)
      File "trainer.py", line 138, in parameter_server
        sess.run(tf.report_uninitialized_variables())
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 929, in run
        run_metadata_ptr)
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1152, in _run
        feed_dict_tensor, options, run_metadata)
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1328, in _do_run
        run_metadata)
      File "/home/experiment/huqiu/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1348, in _do_call
        raise type(e)(node_def, op, message)
    tensorflow.python.framework.errors_impl.UnavailableError: OS Error

    如上所示，运行多机分布式 tensorflow 的 parameter server 进程时，出现这个错误。
    这里说道：

    This has been troubling me for a while. I found out that the problem
    is that GRPC uses the native “epoll” polling engine for communication.
    Changing this to a portable polling engine solved this issue for me.
    The way to do is to set the environment variable,
    “GRPC_POLL_STRATEGY=poll” before running the tensorflow programs. This
    solved this issue for me. For reference, see,
    https://github.com/grpc/grpc/blob/master/doc/environment_variables.md.

    按照其所属，在环境变量中新增一条：

    1
    export GRPC_POLL_STRATEGY=poll

    成功解决问题。

    参考文献