多处理：如何在使用 pool.map 时为每个实例编写单独的日志文件？

Question

我想创建一个 class，每个实例都在其中写入自己的日志文件。当我使用函数而不是 class 时（或者当我不使用多处理时），这工作正常：

import multiprocessing, logging

def setup_logger(name_logfile, path_logfile):
        logger = logging.getLogger(name_logfile)
        formatter = logging.Formatter('%(asctime)s:   %(message)s', datefmt='%Y/%m/%d %H:%M:%S')
        fileHandler = logging.FileHandler(path_logfile, mode='w')
        fileHandler.setFormatter(formatter)
        streamHandler = logging.StreamHandler()
        streamHandler.setFormatter(formatter)

        logger.setLevel(logging.DEBUG)
        logger.addHandler(fileHandler)
        logger.addHandler(streamHandler)
        return logger

    def MyFunc(A):
        print A
        logger = setup_logger('Logfile%s' %A, '/dev/shm/Logfile%s.log' %A)
        logger.info('text to be written to logfile')


    pool = multiprocessing.Pool(2)
    pool.map(MyFunc,[1,2])
    pool.close()
    pool.join()

但是当我使用 class 时，出现酸洗错误：

import multiprocessing, logging

class MyClass(object):
    def __init__(self,A):
        print A
        self.logger = self.setup_logger('Logfile%s' %A, '/dev/shm/Logfile%s.log' %A)
        self.logger.info('text to be written to logfile')

    def setup_logger(self,name_logfile, path_logfile):
        logger = logging.getLogger(name_logfile)
        formatter = logging.Formatter('%(asctime)s:   %(message)s', datefmt='%Y/%m/%d %H:%M:%S')
        fileHandler = logging.FileHandler(path_logfile, mode='w')
        fileHandler.setFormatter(formatter)
        streamHandler = logging.StreamHandler()
        streamHandler.setFormatter(formatter)

        logger.setLevel(logging.DEBUG)
        logger.addHandler(fileHandler)
        logger.addHandler(streamHandler)
        return logger

pool = multiprocessing.Pool(2)
pool.map(MyClass,[1,2])
pool.close()
pool.join()

输出：

1
2
2015/02/12 14:05:09:   text to be written to logfile
2015/02/12 14:05:09:   text to be written to logfile
Process PoolWorker-1:
Traceback (most recent call last):
  File "/usr/lib64/python2.7/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/usr/lib64/python2.7/multiprocessing/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib64/python2.7/multiprocessing/pool.py", line 99, in worker
Process PoolWorker-2:
    put((job, i, result))
  File "/usr/lib64/python2.7/multiprocessing/queues.py", line 392, in put
Traceback (most recent call last):
  File "/usr/lib64/python2.7/multiprocessing/process.py", line 258, in _bootstrap
    return send(obj)
PicklingError: Can't pickle <type 'thread.lock'>: attribute lookup thread.lock failed
    self.run()
  File "/usr/lib64/python2.7/multiprocessing/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/lib64/python2.7/multiprocessing/pool.py", line 99, in worker
    put((job, i, result))
  File "/usr/lib64/python2.7/multiprocessing/queues.py", line 392, in put
    return send(obj)
PicklingError: Can't pickle <type 'thread.lock'>: attribute lookup thread.lock failed

我不知道这个错误的原因是什么，因为每个日志文件都有自己的输出路径。我需要记录器作为对象的一个属性，那么我该如何解决这个 pickling 错误？

Answer 1

基本上，您想调用 multiprocessing.get_logger() 而不是 logging.getLogger()。

查看Python multiprocessing - logging.FileHandler object raises PicklingError

的第一个答案

Answer 2

你不能 pickle 记录器。您可以做的是在对象被 pickled 和 unpickled 时删除并重置记录器：

import multiprocessing, logging


class MyClass(object):

   def __init__(self,A):
        print A
        self.A = A # we need to keep the name!
        self.logger = self.setup_logger('Logfile%s' %A, '/misc/hy5/scheffler/Skripte_Models/python/Tests/Logfile%s.log' %A)
        self.logger.info('text to be written to logfile')

    def setup_logger(self,name_logfile, path_logfile):
        logger = logging.getLogger(name_logfile)
        formatter = logging.Formatter('%(asctime)s:   %(message)s', datefmt='%Y/%m/%d %H:%M:%S')
        fileHandler = logging.FileHandler(path_logfile, mode='w')
        fileHandler.setFormatter(formatter)
        streamHandler = logging.StreamHandler()
        streamHandler.setFormatter(formatter)

        logger.setLevel(logging.DEBUG)
        logger.addHandler(fileHandler)
        logger.addHandler(streamHandler)
        return logger

    def __getstate__(self):
        """Called for pickling.

        Removes the logger to allow pickling and returns a copy of `__dict__`.

        """
        statedict = self.__dict__.copy()
        if 'logger' in statedict:
            # Pickling does not work with loggers objects, so we just keep the logger's name:
            del statedict['logger']
        return statedict

    def __setstate__(self, statedict):
        """Called after loading a pickle dump.

        Restores `__dict__` from `statedict` and adds a new logger.

        """
        self.__dict__.update(statedict)
        process_name = multiprocessing.current_process().name
        self.logger = self.setup_logger('Logfile%s' % self.A, 
                       '/dev/shm/Logfile%s_%s.log' % (self.A, process_name)

请注意，我们将进程名称添加到日志文件中以避免多个进程操作同一个文件！您可能还想确保日志记录处理程序和相应的文件在某个时候关闭。

编辑：

多处理模块中有一个multiprocessing aware logger。但是，我总是觉得这个太局限了。

多处理：如何在使用 pool.map 时为每个实例编写单独的日志文件？

Multiprocessing: How to write separate log files for each instance while using pool.map?

python

class

logging

multiprocessing

instance