python 使用multiprocessing需要注意的问题

我们在编写程序的时候经常喜欢这样写代码

import MySQLdb
import time
from multiprocessing import Process

conn = MySQLdb.connect(‘localhost‘, ‘vearne‘, ‘xx‘, ‘test‘)

def f(name):
    for i in xrange(10):
        cursor = conn.cursor()
        sql = "insert into car(name) values(%s)"
        param = [(name)]
        print param
        #time.sleep(1)
        n = cursor.execute(sql,param)
        cursor.close()
        conn.commit()

if __name__ == ‘__main__‘:
    for i in xrange(10):
        p = Process(target=f, args=(‘bob‘,))
        p.start()

上面的程序有问题吗?
以上的程序在单进程的情况下，应该是没有问题，但是在多进程的情况下，它是有错误的。

首先看看下面的源码

class Process(object):
    ‘‘‘
    Process objects represent activity that is run in a separate process

    The class is analagous to `threading.Thread`
    ‘‘‘
    _Popen = None

    def __init__(self, group=None, target=None, name=None, args=(), kwargs={}):
        assert group is None, ‘group argument must be None for now‘
        count = _current_process._counter.next()
        self._identity = _current_process._identity + (count,)
        self._authkey = _current_process._authkey
        self._daemonic = _current_process._daemonic
        self._tempdir = _current_process._tempdir
        self._parent_pid = os.getpid()
        self._popen = None
        self._target = target
        self._args = tuple(args)
        self._kwargs = dict(kwargs)
        self._name = name or type(self).__name__ + ‘-‘ +                      ‘:‘.join(str(i) for i in self._identity)

    def run(self):
        ‘‘‘
        Method to be run in sub-process; can be overridden in sub-class
        ‘‘‘
        if self._target:
            self._target(*self._args, **self._kwargs)

    def start(self):
        ‘‘‘
        Start child process
        ‘‘‘
        assert self._popen is None, ‘cannot start a process twice‘
        assert self._parent_pid == os.getpid(),                ‘can only start a process object created by current process‘
        assert not _current_process._daemonic,                ‘daemonic processes are not allowed to have children‘
        _cleanup()
        if self._Popen is not None:
            Popen = self._Popen
        else:
            from .forking import Popen
        self._popen = Popen(self)   # -- 创建 Popen 对象 --
        _current_process._children.add(self)
        #  省略部分代码 ... ...
    def _bootstrap(self):    # -- _bootstrap 函数 --
        from . import util
        global _current_process

        try:
            self._children = set()
            self._counter = itertools.count(1)
            try:
                sys.stdin.close()
                sys.stdin = open(os.devnull)
            except (OSError, ValueError):
                pass
            _current_process = self
            util._finalizer_registry.clear()
            util._run_after_forkers()
            util.info(‘child process calling self.run()‘)
            try:
                self.run()  # -- 调用run函数 --
                exitcode = 0
            finally:
                util._exit_function()
        except SystemExit, e:
            if not e.args:
                exitcode = 1
            elif isinstance(e.args[0], int):
                exitcode = e.args[0]
            else:
                sys.stderr.write(str(e.args[0]) + ‘\n‘)
                sys.stderr.flush()
                exitcode = 0 if isinstance(e.args[0], str) else 1
        except:
            exitcode = 1
            import traceback
            sys.stderr.write(‘Process %s:\n‘ % self.name)
            sys.stderr.flush()
            traceback.print_exc()

        util.info(‘process exiting with exitcode %d‘ % exitcode)
        return exitcode

from .forking import Popen 定义

    class Popen(object):

        def __init__(self, process_obj):
            sys.stdout.flush()
            sys.stderr.flush()
            self.returncode = None

            self.pid = os.fork()     # -- fork子进程 --
            # fork 函数调用一次，会在返回两次，一次在父进程中返回，返回的pid 值大于0      
            # 一次在子进程中返回，返回的pid值 等于 0
            if self.pid == 0:        # pid值 等于 0 说明 以下代码都是在子进程中执行的 
                if ‘random‘ in sys.modules:
                    import random
                    random.seed()
                code = process_obj._bootstrap() # -- 调用_bootstrap函数 --
                sys.stdout.flush()
                sys.stderr.flush()
                os._exit(code)

从代码中我们可以看出，python 的multiprocessing 使用fork创建子进程，并在子进程中执行run函数

man fork 

可以得到如下信息

Fork() causes creation of a new process.  The new process (child process) is an exact copy of the calling process (parent process) except for the following:
           o   The child process has a unique process ID.
           o   The child process has a different parent process ID (i.e., the process ID of the parent process).
           o   The child process has its own copy of the parent‘s descriptors.  These descriptors reference the same underlying objects, so that, for instance, file pointers in file objects are shared between the child and the parent, so that an lseek(2) on a descriptor in the child process can affect a subsequent read or write by the parent.  This descriptor copying is also used by the shell to establish standard input and output for newly created processes as well as to set up pipes.
           o   The child processes resource utilizations are set to 0; see setrlimit(2).

fork 函数创建的子进程是父进程的完全拷贝，他们拥有相同的文件描述符。这样如果在父进程中创建了连接，就会出现父进程和多个子进程公用一个连接，会出现无法预料的错误。
（每个连接都有独立的读缓冲区和写缓冲区，多个进程的对读缓冲区和写缓冲区的操作会导致数据混乱）

所以我们应该在子进程中创建连接，这样就能够避免问题的发生。

import MySQLdb
import time
from multiprocessing import Process

class SLWorker(Process):
    def __init__(self):
        super(SLWorker, self).__init__()
        self.conn = None

    def run(self):
        # *** 注意这里 *** 连接延迟加载, 也就是说连接在子进程中被创建
        if self.conn ==  None:  
            self.conn = MySQLdb.connect(‘localhost‘, ‘vearne‘, ‘xxx‘, ‘test‘)
        for i in xrange(10):
            cursor = self.conn.cursor()
            sql = "insert into car(name) values(%s)"
            name = "bob"
            param = [(name)]
            print param
            #time.sleep(30)
            n = cursor.execute(sql,param)
            cursor.close()
            self.conn.commit()
    def __del__(self):
        if self.conn != None:
            self.conn.close()

if __name__ == ‘__main__‘:
    ll = []
    for i in xrange(10):
        p = SLWorker()
        p.start()
        ll.append(p)
    for p in ll:
        p.join()

答案归结为只需要将在子进程中创建连接，或者连接延迟创建就能够解决这个问题
其实现在有很多连接池都是延迟创建连接，没有仔细看，有研究的分享下。

PS: celery 和rq 也都会有这样的问题，请大家引起足够重视