Python logging: subloggers or extra information

Question

I have a Python application that uses dynamically discovered plugins to extends it capabilities. I am trying to figure out what is the best way to log information from the plugins. I think there are at least two solutions. Either I create a hierarchy of loggers:

myapp
myapp.plugin1
myapp.plugin2

And withing the base app:

logger = logging.getLogger('myapp')
logger.debug('mymessage')

and withing the plugin1

logger = logging.getLogger('myapp.plugin1')
logger.debug('mymessage from plugin1')

Or I add additional information using the extra parameter in the default logger: Within base app:

logger = logging.getLogger('myapp')
logger.debug('mymessage', extra=dict(plugin=None)

Within plugin1

logger = logging.getLogger('myapp')
logger.debug('mymessage from plugin1', extra=dict(plugin='plugin1')

What are the advantages or disadvantages (if any) between these two schemes?

Answer 1

The best practice in the Python world is that each Python module declares its own logger. At the head of each file you have:

 import logging

 logger = logging.getLogger(__name__)

Then you have your Python modules in nice hierarchy, so that you can tell to increase or decrease the package logging level hierarchically using the dotted package name lookups. Some examples:

    'django': {
        'handlers': ['mail_admins'],
        'level': 'WARN',
        'propagate': True,
    },


    'django.request': {
        'handlers': ['mail_admins'],
        'level': 'ERROR',
        'propagate': True,
    },

    'django.db.backends': {
        'handlers': [],
        'level': 'ERROR',
        'propagate': True,
    },