在实际使用多进程的过程中，希望一个子进程报错，则停止所有进程，并退出主进程。在子进程中不能使用exit()函数，它会使得主进程一直等待，程序就卡在这里了。
代码语言：javascript复制from multiprocessing import Pool
import time
import os
import sys
import signal
import subprocess
import platform
def f(i):
print(f'running {i}, pid: {os.getgid()}')
if i == 3:
time.sleep(5)
print('exit in subp')
exit()
time.sleep(3)
return i
def call_back(i):
print('callback:',i)
def error_call_back(i):
print('error:',i)
def main():
mp = Pool(4)
for i in range(10):
mp.apply_async(f,args=(i,),callback=call_back,error_callback=error_call_back)
mp.close()
mp.join()
if __name__ == '__main__':
main()程序运行：
代码语言：javascript复制running 0, pid: 1000
running 1, pid: 1000
running 2, pid: 1000
running 3, pid: 1000
callback: 0
running 4, pid: 1000
callback: 1
running 5, pid: 1000
callback: 2
running 6, pid: 1000
exit in subp
None
running 7, pid: 1000
callback: 4
running 8, pid: 1000
callback: 5
running 9, pid: 1000
callback: 6
callback: 7
callback: 8
callback: 9程序并不会退出。
解决方案：使用在error_callback中使用killpg函数，杀死进程组。
代码语言：javascript复制from multiprocessing import Pool
import time
import os
import sys
import signal
import subprocess
import platform
def f(i):
print(f'running {i}, pid: {os.getgid()}')
if i == 3:
time.sleep(5)
print('subprocess raise error, jump to error callback')
raise ValueError(f'error in {i}')
time.sleep(3)
return i
def call_back(i):
print('callback:',i)
def error_call_back(i):
print('error:',i)
print('exit by killpg')
kill_sig = signal.SIGKILL
os_name = platform.system()
if os_name == 'Windows':
kill_sig = signal.CTRL_C_EVENT
os.killpg(os.getpgid(os.getpid()), kill_sig)
def main():
mp = Pool(4)
for i in range(10):
mp.apply_async(f,args=(i,),callback=call_back,error_callback=error_call_back)
mp.close()
mp.join()
if __name__ == '__main__':
main()运行结果:
代码语言：javascript复制running 0, pid: 1000
running 1, pid: 1000
running 2, pid: 1000
running 3, pid: 1000
callback: 0
running 4, pid: 1000
callback: 1
running 5, pid: 1000
callback: 2
running 6, pid: 1000
subprocess raise error, jump to error callback
running 7, pid: 1000
error: error in 3
exit by killpg这样就能解决两个问题： （1）子进程发生异常时所有子进程退出且主进程退出。 （2）子进程中不能使用exit()函数问题。
