python multiprocessing pool retries

multiprocessing, python

Solution

If you can (or don't mind) retrying immediately, use a decorator wrapping the function:

import random
from multiprocessing import Pool
from functools import wraps

def retry(f):
    @wraps(f)
    def wrapped(*args, **kwargs):
        while True:
            try:
                return f(*args, **kwargs)
            except ValueError:
                pass
    return wrapped

@retry
def f(x):
    if random.getrandbits(1):
        raise ValueError("Retry this computation")
    return x*x

p = Pool(5)
# If one of these f(x) calls fails, retry it with another (or same) process
p.map(f, [1,2,3])

Problem

Original source