python僵尸进程产生的原因

在 unix 或 unix-like 的系统中，当一个子进程退出后，它就会变成一个僵尸进程，如果父进程没有通过 wait 系统调用来读取这个子进程的退出状态的话，这个子进程就会一直维持僵尸进程状态。

Zombie process - Wikipedia 中是这样描述的：

On Unix and Unix-like computer operating systems, a zombie process or defunct process is a process that has completed execution (via the exit system call) but still has an entry in the process table: it is a process in the "Terminated state". This occurs for child processes, where the entry is still needed to allow the parent process to read its child's exit status: once the exit status is read via the wait system call, the zombie's entry is removed from the process table and it is said to be "reaped". A child process always first becomes a zombie before being removed from the resource table. In most cases, under normal system operation zombies are immediately waited on by their parent and then reaped by the system – processes that stay zombies for a long time are generally an error and cause a resource leak.

并且僵尸进程无法通过 kill 命令来清除。

本文将探讨如何手动制造一个僵尸进程以及清除僵尸进程的办法。

手动制造一个僵尸进程

为了便于后面讲解清除僵尸进程的方法，我们使用日常开发中经常使用的 multiprocessing 模块来制造僵尸进程（准确的来说是制造一个长时间维持僵尸进程状态的子进程）：

$ cat test_a.py
from multiprocessing import Process, current_process
import logging
import os
import time

logging.basicConfig(
  level=logging.DEBUG,
  format='%(asctime)-15s - %(levelname)s - %(message)s'
)

def run():
  logging.info('exit child process %s', current_process().pid)
  os._exit(3)

p = Process(target=run)
p.start()
time.sleep(100)

测试：

$ python test_a.py &
[1] 10091
$ 2017-07-20 21:28:14,792 - INFO - exit child process 10106

$ ps aux |grep 10106
mozillazg       10126  0.0 0.0 2434836  740 s006 R+  0:00.00 grep 10106
mozillazg       10106  0.0 0.0    0   0 s006 Z   0:00.00 (Python)

可以看到，子进程 10091 变成了僵尸进程。

既然已经可以控制僵尸进程的产生了，那我们就可以进入下一步如何清除僵尸进程了。

清除僵尸进程有两种方法：

•第一种方法就是结束父进程。当父进程退出的时候僵尸进程随后也会被清除。
• 第二种方法就是通过 wait 调用来读取子进程退出状态。我们可以通过处理 SIGCHLD 信号，在处理程序中调用 wait 系统调用来清除僵尸进程。

处理 SIGCHLD 信号

子进程退出时系统会向父进程发送 SIGCHLD 信号，父进程可以通过注册 SIGCHLD 信号处理程序，在信号处理程序中调用 wait
系统调用来清理僵尸进程。 $ cat test_b.py

import errno
from multiprocessing import Process, current_process
import logging
import os
import signal
import time

logging.basicConfig(
  level=logging.DEBUG,
  format='%(asctime)-15s - %(levelname)s - %(message)s'
)

def run():
  exitcode = 3
  logging.info('exit child process %s with exitcode %s',
         current_process().pid, exitcode)
  os._exit(exitcode)

def wait_child(signum, frame):
  logging.info('receive SIGCHLD')
  try:
    while True:
      # -1 表示任意子进程
      # os.WNOHANG 表示如果没有可用的需要 wait 退出状态的子进程，立即返回不阻塞
      cpid, status = os.waitpid(-1, os.WNOHANG)
      if cpid == 0:
        logging.info('no child process was immediately available')
        break
      exitcode = status >> 8
      logging.info('child process %s exit with exitcode %s', cpid, exitcode)
  except OSError as e:
    if e.errno == errno.ECHILD:
      logging.error('current process has no existing unwaited-for child processes.')
    else:
      raise
  logging.info('handle SIGCHLD end')

signal.signal(signal.SIGCHLD, wait_child)

p = Process(target=run)
p.start()

while True:
  time.sleep(100)

效果：

$ python test_b.py &
[1] 10159
$ 2017-07-20 21:28:56,085 - INFO - exit child process 10174 with exitcode 3
2017-07-20 21:28:56,088 - INFO - receive SIGCHLD
2017-07-20 21:28:56,089 - INFO - child process 10174 exit with exitcode 3
2017-07-20 21:28:56,090 - ERROR - current process has no existing unwaited-for child processes.
2017-07-20 21:28:56,090 - INFO - handle SIGCHLD end

$ ps aux |grep 10174
mozillazg       10194  0.0 0.0 2432788  556 s006 R+  0:00.00 grep 10174

可以看到，子进程退出变成僵尸进程后，系统给父进程发送了 SIGCHLD 信号，我们在 SIGCHLD 信号的处理程序中通过 os.waitpid 调用 wait 系统调用后阻止了子进程一直处于僵尸进程状态，从而实现了清除僵尸进程的效果。