Multiprocessing Pool在Ubuntu上运行卡顿，Windows正常的问题求助

问题现象 #

• Windows系统中脚本运行正常，CPU利用率达预期100%，串行版本执行无异常
• 部署到Ubuntu系统后，Multiprocessing Pool完成第一次进程迭代后卡住，CPU利用率处于99.5%的空闲状态

已尝试方案 #

• 设置进程启动方法为spawn：

  mp.set_start_method("spawn")
  

• 配置maxtasksperchild=1参数

次要问题 #

日志模块在Windows和Ubuntu系统下均无法正常工作

启动代码 #

if __name__ == '__main__':
    mp.set_start_method("spawn")
    logging.basicConfig(filename='seasonality.log', level=logging.DEBUG)

    EOD_Hist = eod_historical.PullEOD(apikey, 'json')

    ticker = EOD_Hist.get_index_components_list(index)

    def main(ticker_):
        start_year = 1970
        start = '2023-01-01'
        end = '2023-12-31'
        logging.debug(f'{datetime.now()}: {ticker_} FULL TASKS STARTED - params:  start_year = {start_year} / start = {start} / end = {end}')
        hist, iter_range = raw_data(start_year, ticker_)
        start_df, end_df, performance_df, average_perf_df, median_perf_df, perf_stdv, max_draw_down_df, hit_ratio = period_iter(start, end, hist, iter_range)
        df = create_df(start_df, end_df, performance_df, average_perf_df, median_perf_df, perf_stdv, max_draw_down_df, hit_ratio)
        helper_functions.write_df_to_database(mongoip, mongouser, mongopw, mongotable, ticker, df)
        logging.debug(f'{datetime.now()}: {ticker_} FULL TASKS ENDED - length_df_written = {df}')
        return f'{ticker} done'

    pool = mp.Pool(os.cpu_count(), maxtasksperchild=1)                         # Create a multiprocessing Pool
    result = pool.map(main, ticker)
    print(result)

补充说明 #

被调用的函数运行正常，串行版本可正常执行，且API调用的每日额度在正常减少（说明API请求逻辑无问题）

内容的提问来源于stack exchange，提问作者Lorenz Weidinger