“多处理”；叉子；使用Python 3.6.1的Linux/Intel Xeon上的上下文块？

问题描述
我稍微调整了代码（见下文）。但是，当在Linux上运行此脚本时（因此命令行：

python script\u name.py

），它将为所有作业打印

jobs running:x

，但之后似乎就卡住了。但是，当我使用spawn方法（

mp.set_start_方法（'spawn'）

）时，它工作正常，并立即开始打印

计数器

变量的值（请参见

侦听器

方法）

问题

Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                20
On-line CPU(s) list:   0-19
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             20
NUMA node(s):          2
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 45
Model name:            Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz
Stepping:              2
CPU MHz:               2596.501
BogoMIPS:              5193.98
Hypervisor vendor:     VMware
Virtualization type:   full
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              25600K
NUMA node0 CPU(s):     0-19

• 为什么它只在生成进程时工作
• 如何调整代码，使其与

  fork

  配合使用？（因为它可能更快）

代码

import io
import csv
import multiprocessing as mp

NEWLINE = '\n'

def file_searcher(file_path):
    parsed_file = csv.DictReader(io.open(file_path, 'r', encoding='utf-8'), delimiter='\t')

    manager = mp.Manager()
    q = manager.Queue()
    pool = mp.Pool(mp.cpu_count())

    # put listener to work first
    watcher = pool.apply_async(listener, (q,))

    jobs = []
    for row in parsed_file:
        print('jobs running: ' + str(len(jobs) + 1))
        job = pool.apply_async(worker, (row, q))
        jobs.append(job)

  # collect results from the workers through the pool result queue
    for job in jobs:
        job.get()

    #now we are done, kill the listener
    q.put('kill')
    pool.close()
    pool.join()

def worker(genome_row, q):
    complete_data = []
    #data processing
    #ftp connection to retrieve data
    #etc.
    q.put(complete_data)
    return complete_data

def listener(q):
    '''listens for messages on the q, writes to file. '''
    f = io.open('output.txt', 'w', encoding='utf-8')
    counter = 0
    while 1:
        m = q.get()
        counter +=1
        print(counter)
        if m == 'kill':
            break
        for x in m:
            f.write(x + NEWLINE)
        f.flush()
    f.close()

if __name__ == "__main__":
   file_searcher('path_to_some_tab_del_file.txt')

处理器信息

Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                20
On-line CPU(s) list:   0-19
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             20
NUMA node(s):          2
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 45
Model name:            Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz
Stepping:              2
CPU MHz:               2596.501
BogoMIPS:              5193.98
Hypervisor vendor:     VMware
Virtualization type:   full
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              25600K
NUMA node0 CPU(s):     0-19

Linux内核版本

3.10.0-514.26.2.el7.x86_64

Python版本

Python 3.6.1 :: Continuum Analytics, Inc.

---

日志
我按照@yacc的建议添加了代码，这将给出以下日志：

[server scripts]$ python main_v3.py
[INFO/SyncManager-1] child process calling self.run()
[INFO/SyncManager-1] created temp directory /tmp/pymp-2a9stjh6
[INFO/SyncManager-1] manager serving at '/tmp/pymp-2a9stjh6/listener-jxwseclw'
[DEBUG/MainProcess] requesting creation of a shared 'Queue' object
[DEBUG/SyncManager-1] 'Queue' callable returned object with id '7f0842da56a0'
[DEBUG/MainProcess] INCREF '7f0842da56a0'
[DEBUG/MainProcess] created semlock with handle 139673691570176
[DEBUG/MainProcess] created semlock with handle 139673691566080
[DEBUG/MainProcess] created semlock with handle 139673691561984
[DEBUG/MainProcess] created semlock with handle 139673691557888
[DEBUG/MainProcess] added worker
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-2] INCREF '7f0842da56a0'
[DEBUG/MainProcess] added worker
[INFO/ForkPoolWorker-2] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-4] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-4] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-3] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-3] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-6] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-5] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-6] child process calling self.run()
[INFO/ForkPoolWorker-5] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-7] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-8] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-7] child process calling self.run()
[INFO/ForkPoolWorker-8] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-9] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-9] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-10] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-10] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-11] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-11] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-12] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-12] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-13] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-13] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-14] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-14] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-15] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-15] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-16] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-16] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-17] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-17] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-18] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-18] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-19] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-19] child process calling self.run()
[DEBUG/MainProcess] added worker
[DEBUG/ForkPoolWorker-20] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-20] child process calling self.run()
jobs running: 1
jobs running: 2
jobs running: 3
jobs running: 4
[DEBUG/ForkPoolWorker-21] INCREF '7f0842da56a0'
[INFO/ForkPoolWorker-21] child process calling self.run()
jobs running: 5
jobs running: 6
jobs running: 7
[DEBUG/ForkPoolWorker-2] INCREF '7f0842da56a0'
jobs running: 8
written to file
jobs running: 9
jobs running: 10
[DEBUG/ForkPoolWorker-2] thread 'MainThread' does not own a connection
[DEBUG/ForkPoolWorker-2] making connection to manager
jobs running: 11
jobs running: 12
jobs running: 13
jobs running: 14
jobs running: 15
[DEBUG/SyncManager-1] starting server thread to service 'ForkPoolWorker-2'
jobs running: 16
jobs running: 17
jobs running: 18
jobs running: 19
[DEBUG/ForkPoolWorker-4] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-3] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-5] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-6] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-7] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-8] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-10] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-9] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-11] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-13] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-14] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-12] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-15] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-16] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-18] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-17] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-20] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-19] INCREF '7f0842da56a0'
[DEBUG/ForkPoolWorker-21] INCREF '7f0842da56a0'

---

正如@jxh所暗示的，fork和spawn之间的差异非常重要。在第17.2.1.2节中指出，区别在于：forking保留了环境和stdin/out之类的东西，而spawn只是创建了一个新的过程。我认为您的环境中可能存在导致worker函数出现问题的内容，可能是您对其他处理的注释背后的代码。产卵给你一个干净的板岩，事情在这些条件下运行良好

为了确定发生了什么，我会让每个工作人员打印诊断消息，可能会写入每个工作人员特有的文件中。每次要编写消息时打开/关闭该文件，以便更新/刷新内容

---

Fork不应该比spawn快，因为Fork需要将环境信息复制到新进程。在任何情况下，这只是启动成本，我认为这是最小的，因为工作者需要做一些计算或I/O绑定的工作，您希望并行化

你能提供关于Linux、Python、MP软件包和硬件/处理器版本的详细信息吗？谢谢，至少我的问题现在更具体了，因为我知道你的日志@yaccIt有点尴尬，但至少你有一个关于繁殖的工作。我建议升级到3.6.2，如果仍然出现这种情况，请在Python上打开一张罚单。这对于未来版本的修复可能很重要。这意味着你的“机器”实际上是虚拟的，而不是裸机。这意味着存在一个硬件抽象层，它假装是Linux的机器，而实际上是一个运行在主机操作系统上的进程

VMware

是一家创建虚拟机监控程序的软件公司，虚拟机监控程序是在主机操作系统上运行的软件，用于向虚拟机提供虚拟硬件。由于

fork

必须共享来自父系统的资源，Linux很可能会设置锁。在真正的硬件上，这个锁会很便宜，但在虚拟机上，不清楚这个锁的虚拟版本会有多便宜。