我更喜欢使用配置文件。它允许我在从开发到发布的过程中切换日志级别、位置等，而无需更改代码。我只是用相同的名称打包了一个不同的配置文件，并定义了相同的记录器。

import logging.config
if __name__ == '__main__':
    # Configure the logger
    # loggerConfigFileName: The name and path of your configuration file
    logging.config.fileConfig(path.normpath(loggerConfigFileName))

    # Create the logger
    # Admin_Client: The name of a logger defined in the config file
    mylogger = logging.getLogger('Admin_Client')

    msg='Bite Me'
    myLogger.debug(msg)
    myLogger.info(msg)
    myLogger.warn(msg)
    myLogger.error(msg)
    myLogger.critical(msg)

    # Shut down the logger
    logging.shutdown()

下面是日志配置文件的代码

#These are the loggers that are available from the code
#Each logger requires a handler, but can have more than one
[loggers]
keys=root,Admin_Client


#Each handler requires a single formatter
[handlers]
keys=fileHandler, consoleHandler


[formatters]
keys=logFormatter, consoleFormatter


[logger_root]
level=DEBUG
handlers=fileHandler


[logger_Admin_Client]
level=DEBUG
handlers=fileHandler, consoleHandler
qualname=Admin_Client
#propagate=0 Does not pass messages to ancestor loggers(root)
propagate=0


# Do not use a console logger when running scripts from a bat file without a console
# because it hangs!
[handler_consoleHandler]
class=StreamHandler
level=DEBUG
formatter=consoleFormatter
args=(sys.stdout,)# The comma is correct, because the parser is looking for args


[handler_fileHandler]
class=FileHandler
level=DEBUG
formatter=logFormatter
# This causes a new file to be created for each script
# Change time.strftime("%Y%m%d%H%M%S") to time.strftime("%Y%m%d")
# And only one log per day will be created. All messages will be amended to it.
args=("D:\\Logs\\PyLogs\\" + time.strftime("%Y%m%d%H%M%S")+'.log', 'a')


[formatter_logFormatter]
#name is the name of the logger root or Admin_Client
#levelname is the log message level debug, warn, ect 
#lineno is the line number from where the call to log is made
#04d is simple formatting to ensure there are four numeric places with leading zeros
#4s would work as well, but would simply pad the string with leading spaces, right justify
#-4s would work as well, but would simply pad the string with trailing spaces, left justify
#filename is the file name from where the call to log is made
#funcName is the method name from where the call to log is made
#format=%(asctime)s | %(lineno)d | %(message)s
#format=%(asctime)s | %(name)s | %(levelname)s | %(message)s
#format=%(asctime)s | %(name)s | %(module)s-%(lineno) | %(levelname)s | %(message)s
#format=%(asctime)s | %(name)s | %(module)s-%(lineno)04d | %(levelname)s | %(message)s
#format=%(asctime)s | %(name)s | %(module)s-%(lineno)4s | %(levelname)-8s | %(message)s

format=%(asctime)s | %(levelname)-8s | %(lineno)04d | %(message)s


#Use a separate formatter for the console if you want
[formatter_consoleFormatter]
format=%(asctime)s | %(levelname)-8s | %(filename)s-%(funcName)s-%(lineno)04d | %(message)s

最好干净的方法。

    ds = datetime.now().strftime("%Y%m%d_%H%M%S")
    try:
        # py39 有force参数指定可能强制除去之前的handler，这里使用兼容写法，0708
        logging.getLogger().removeHandler(logging.getLogger().handlers[0])
        logging.getLogger().removeHandler(logging.getLogger().handlers[0])
    except:
        pass
    logging.basicConfig(
        # force=
        level=logging.INFO,
        format="%(asctime)s [%(levelname)s] %(message)s",
        handlers=[
            logging.FileHandler('log_%s_%s_%s.log' % (ds, create_net, os.path.basename(dataset_path))),
            logging.StreamHandler(sys.stdout)
        ]
    )
    logging.info('start train')

http://docs.python.org/library/logging.handlers.html#filehandler

FileHandler类位于核心日志包中，它将日志输出发送到磁盘文件。

虽然这是一个老问题，但对于现在遇到这个问题的人来说，您也可以使用dictConfig。例如，对于info级别以上的文件:

logging.config.dictConfig({
    'version': 1,
    'formatters': {
        'default': {
            'format': '[%(asctime)s] %(message)s',
        }
    },
    'handlers': {
        'info': {
            'level': logging.INFO,
            'class': 'logging.FileHandler',
            'filename': 'info.log',
        },
    },
    "root": {
        "level": logging.INFO,
        "handlers": ["info"]
    }
})

或者另一个更具体的例子，在一个特定的目录下旋转文件:

today = datetime.date.today()
folder = './log'
Path(folder).mkdir(parents=True, exist_ok=False) # Create folder if not exists
logging.config.dictConfig({
        ...
        'info': {
            'level': logging.INFO,
            'class': 'logging.handlers.TimedRotatingFileHandler',
            'filename': f'{folder}/info-{today.month:02}-{today.year}.log',
            # Roll over on the first day of the weekday
            'when': 'W0',
            # Roll over at midnight
            'atTime': datetime.time(hour=0),
            # Number of files to keep.
            'backupCount': 8
        },
        ...

http://docs.python.org/library/logging.html#logging.basicConfig

logging.basicConfig(filename='/path/to/your/log', level=....)

摘自“伐木食谱”:

# create logger with 'spam_application'
logger = logging.getLogger('spam_application')
logger.setLevel(logging.DEBUG)
# create file handler which logs even debug messages
fh = logging.FileHandler('spam.log')
fh.setLevel(logging.DEBUG)
logger.addHandler(fh)

你可以开始了。

附注:一定要阅读日志指南。