Python 在不同的日志级别记录到多个处理程序？

Question

各位，

我正在为一个我无法正确处理的 python 日志配置而挠头。

假设我安装了以下软件包：

mypackage/
   data/mypackage.logging.conf
   module1.py
   module2.py
   scripts/main.py

由于脚本可以交互使用或从 crontab 运行，我有以下要求：

1. 没有打印语句，一切都经过日志记录；

2. 使用timedRotatingFileHandler记录日志，始终处于调试级别；

3. 使用mailinglogger.SummarisingLogger ，总是在 INFO 级别；

4. 登录到控制台，级别默认设置为 INFO 或通过命令行选项覆盖。

问题是，我可以通过命令行更改日志级别并相应地更改控制台日志级别，但其他处理程序也发生了更改，这是我不想要的... :-/

在日志记录配置文件中，我不确定我是否理解根记录器的级别、其他记录器的级别和处理程序的级别设置之间的优先级。

这是一些示例代码。 任何线索将不胜感激:-)

# mypackage/data/mypackage.logging.conf
[loggers]
root,mypackage

[handlers]
keys=consoleHandler,timedRotatingFileHandler,summarisingHandler

[formatters]
keys=simpleFormatter,consoleFormatter,mypackageFormatter

[logger_root]
#level=INFO
handlers=consoleHandler

[logger_mypackage]
#level=INFO
handlers=timedRotatingFileHandler,summarisingHandler
qualname=mypackage

[handler_consoleHandler]
class=StreamHandler
#level=INFO
formatter=consoleFormatter
args=(sys.stdout,)

[handler_timedRotatingFileHandler]
class=logging.handlers.TimedRotatingFileHandler
level=DEBUG
formatter=mypackageFormatter
args=('mypackage.log', 'M', 1, 5)

[handler_summarisingHandler]
class=mailinglogger.SummarisingLogger
level=INFO
formatter=mypackageFormatter
args=('mypackage@someserver.somewhere.com', ('mypackage-alerts@somewhere.com',), 'relay.somewhere.com')

---

#mypackage/scripts/main.py:
import logging
import logging.config
import os
import sys

import mypackage.module1
import mypackage.module2

logging.config.fileConfig('data/mypackage.logging.conf')
log = logging.getLogger(__name__)

if __name__ == '__main__':
    loglevel = 'INFO'
    if len(sys.argv) > 1:
        loglevel = sys.argv[1].upper()

    logging.getLogger('').setLevel(getattr(logging, loglevel))
    # or logging.getLogger('mypackage').setLevel(getattr(logging, loglevel)) ?

    mypackage.module1.do_something()
    mypackage.module2.do_something_else()

---

#mypackage/module1.py:
import logging

log = logging.getLogger(__name__)
log.addHandler(NullHandler())

def do_something():
    log.debug("some debug message from:" + __name__)
    log.info("some info message from:" + __name__)
    log.error("some error message from:" + __name__)

---

#mypackage/module2.py:
import logging

log = logging.getLogger(__name__)
log.addHandler(NullHandler())

def do_something_else():
    log.debug("some debug message from:" + __name__)
    log.info("some info message from:" + __name__)
    log.error("some error message from:" + __name__)

---

更新 1

与此同时，我发现了这个答案并通过这种方式成功修改了我的代码：

#mypackage/scripts/main.py:
import logging
import logging.config
import os
import sys
import mailinglogger

import mypackage.module1
import mypackage.module2

def main():
    # get the console log level from the command-line
    loglevel = 'INFO'
    if len(sys.argv) > 1:
        loglevel = sys.argv[1].upper()

    # create formatters
    simple_formatter = logging.Formatter("%(name)s:%(levelname)s: %(message)s")
    detailed_formatter = logging.Formatter("%(asctime)s %(name)s[%(process)d]: %(levelname)s - %(message)s")

    # get a top-level "mypackage" logger,
    # set its log level to DEBUG,
    # BUT PREVENT IT from propagating messages to the root logger
    #
    log = logging.getLogger('mypackage')
    log.setLevel(logging.DEBUG)
    log.propagate = 0

    # create a console handler
    # and set its log level to the command-line option 
    # 
    console_handler = logging.StreamHandler(sys.stdout)
    console_handler.setLevel(getattr(logging, loglevel))
    console_handler.setFormatter(simple_formatter)

    # create a file handler
    # and set its log level to DEBUG
    #
    file_handler = logging.handlers.TimedRotatingFileHandler('mypackage.log', 'M', 1, 5)
    file_handler.setLevel(logging.DEBUG)
    file_handler.setFormatter(detailed_formatter)

    # create a mail handler
    # and set its log level to INFO
    #
    mail_handler = mailinglogger.SummarisingLogger(
        'mypackage@someserver.somewhere.com', ('mypackage-alerts@somewhere.com',), 'relay.somewhere.com')
    mail_handler.setLevel(logging.INFO)
    mail_handler.setFormatter(detailed_formatter)

    # add handlers to the "mypackage" logger
    #
    log.addHandler(console_handler)
    log.addHandler(file_handler)
    log.addHandler(mail_handler)

    # let the modules do their stuff 
    # and log to the "mypackage.module1" and "mypackage.module2" loggers
    #
    mypackage.module1.do_something()
    mypackage.module2.do_something_else()


if __name__ == '__main__':
    main()

现在，我将尝试在 logging.config 文件中翻译它...

---

更新 2

这是我找到的最好的日志配置和代码组合。

在 mypackage.logging.conf 文件中，“mypackage”记录器是：

• 仅使用文件和电子邮件处理程序进行设置；
• 它的传播设置为false；
• 其级别设置为调试；
• 而文件和电子邮件处理程序分别设置为 INFO 和 DEBUG。

---

#mypackage/data/mypackage.logging.conf
[loggers]
keys=root,mypackage

[handlers]
keys=consoleHandler,timedRotatingFileHandler,summarisingHandler

[formatters]
keys=simpleFormatter,consoleFormatter,mypackageFormatter

[logger_root]
#level=INFO
handlers=consoleHandler

[logger_mypackage]
level=DEBUG
handlers=timedRotatingFileHandler,summarisingHandler
qualname=mypackage
propagate=0

[handler_consoleHandler]
class=StreamHandler
#level=INFO
formatter=consoleFormatter
args=(sys.stdout,)

[handler_timedRotatingFileHandler]
class=logging.handlers.TimedRotatingFileHandler
level=DEBUG
formatter=mypackageFormatter
args=('mypackage.log', 'M', 1, 5)

[handler_summarisingHandler]
class=mailinglogger.SummarisingLogger
level=INFO
formatter=mypackageFormatter
args=('mypackage@someserver.somewhere.com', ('mypackage-alerts@somewhere.com',), 'relay.somewhere.com')

[formatter_consoleFormatter]
format=%(levelname)s: %(message)s
datefmt=

[formatter_mypackageFormatter]
format=%(asctime)s %(name)s[%(process)d]: %(levelname)s - %(message)s
datefmt=

---

在脚本中：

1. 读取日志配置；

2. 一个 console_formatter 是（重新）创建的；

3. 使用命令行选项中的日志级别创建控制台处理程序，然后将其添加到“mypackage”记录器中。

---

import logging
import logging.config
import os
import sys

import mypackage.module1
import mypackage.module2

def setup_logging(loglevel):
    #
    # load logging config from file
    #
    logging.config.fileConfig('data/mypackage.logging.conf')

    # (re-)create formatter
    console_formatter = logging.Formatter("%(name)s:%(levelname)s: %(message)s")

    # create a console handler
    # and set its log level to the command-line option 
    # 
    console_handler = logging.StreamHandler(sys.stdout)
    console_handler.setFormatter(console_formatter)
    console_handler.setLevel(getattr(logging, loglevel))

    # add console handler to the pre-configured "mypackage" logger
    #
    logger = logging.getLogger('mypackage')
    logger.addHandler(console_handler)


def main():
    # get the console log level from the command-line
    loglevel = 'INFO'
    if len(sys.argv) > 1:
        loglevel = sys.argv[1].upper()

    # load logging config and setup console handler
    #
    setup_logging(loglevel)

    # log from the script to the "mypackage.scripts.main" logger
    #
    log = logging.getLogger(__name__)
    log.debug("some debug message from:" + __name__)
    log.info("some info message from:" + __name__)
    log.error("some error message from:" + __name__)

    # let the modules do their stuff 
    # and log to the "mypackage.module1" and "mypackage.module2" loggers
    #
    mypackage.module1.do_something()
    mypackage.module2.do_something_else()

if __name__== '__main__':
    main()

---

如果处理程序在从配置文件加载时按名称“可寻址”，事情会更简单。

然后，我们可以在配置文件中设置 mypackage 控制台处理程序，并在代码中更改其日志级别，如下所示：

def setup_logging(loglevel):
    logging.config.fileConfig('data/mypackage.logging.conf')

    logger = logging.getLogger('mypackage')
    console_handler = logger.getHandler('consoleHandler')
    console_handler.setLevel(getattr(logging, loglevel))

也不需要重新创建格式化程序......

（最后更新：是的，我知道https://docs.python.org/3/library/logging.config.html#incremental-configuration ，但在这种情况下，我坚持使用 Python 2.6 ... :-)

Answer 1

使用 dictConfig。 这是在 dictConfig 中使用单独句柄记录到多个文件的示例。 这并不完全是您要查找的内容，但是您可以修改此示例并简单地更改您要使用的每个处理程序的级别。

import os, logging
from logging.config import dictConfig

FORMAT = "%(asctime)s {app} [%(thread)d] %(levelname)-5s %(name)s - %(message)s. [file=%(filename)s:%(lineno)d]"
DATE_FORMAT = None


def setup_logging(name, level="INFO", fmt=FORMAT):
    formatted = fmt.format(app=name)
    log_dir = r'C:/log_directory'
    if not os.path.exists(log_dir):
        os.makedirs(log_dir)

    logging_config = {
        "version": 1,
        'disable_existing_loggers': False,
        "formatters": {
            'standard': {
                'format': formatted
            }
        },
        "handlers": {
            'default': {
                'class': 'logging.StreamHandler',
                'formatter': 'standard',
                'level': level,
                'stream': 'ext://sys.stdout'
            },
            'file': {
                'class': 'logging.handlers.TimedRotatingFileHandler',
                'when': 'midnight',
                'utc': True,
                'backupCount': 5,
                'level': level,
                'filename': '{}/log1.log'.format(log_dir),
                'formatter': 'standard',
            },
            'file2': {
                'class': 'logging.handlers.TimedRotatingFileHandler',
                'when': 'midnight',
                'utc': True,
                'backupCount': 5,
                'level': level,
                'filename': '{}/log2.log'.format(log_dir),
                'formatter': 'standard',
            }
        },
        "loggers": {
            "": {
                'handlers': ['default', 'file'],
                'level': level
            },
            "second_log": {
                'handlers': ['default', 'file2'],
                'level': level
            }
        }
    }

    dictConfig(logging_config)

log.setup_logging(name="log-name", level=LEVEL
logger = logging.getLogger(__name__)
second_logger = logging.getLogger('second_log')
second_logger.propagate = False

Answer 2

一种更新处理程序的方法：

import logging

from rootmodule.mymodule import mylogger

def update_handler_level(logger,  handler_type, level="INFO"):
    # if not root logger user logger.parent
    for handler in logger.handlers or logger.parent.handlers:
        if isinstance(handler, handler_type):
            print(handler.level)
            handler.setLevel(getattr(logging, level, "INFO"))
            print(handler.level)

mylogger.debug('test')
update_handler_level(mylogger, logging.StreamHandler)
mylogger.debug('test')

我的 logging.cfg 与您的非常相似，除了记录器名称 si 设置在常量模块中（可以在不破坏日志记录配置的情况下进行模块重命名）

要从命令行更新，您的 opts 值和 logging.Handler 子类名称之间必须有一个映射。