Develop Reference

Python root logger messages not being logged via handler configured with fileConfig

The Problem:
Given a logging config and a logger that employs that config, I see log messages from the script in which the log handler is configured, but not from the root logger, to which the same handler is assigned.
Details:
(Using Python 2.7)
I have a module my_mod which instantiates a logger. my_mod has a function my_command which logs some messages using that logger. my_mod exists inside of a library my_lib, so I don't want to configure the logger with any handlers; as recommended, I want to leave the log handling to the developer using my_mod. my_mod looks like:
import logging
LOGGER = logging.getLogger(__name__)
def my_command():
LOGGER.debug("This is a log message from module.py")
print "This is a print statement from module.py"
I also have a python script my_script.py, which uses my_mod.my_command. my_script.py instantiates a logger, and in this case I do have handlers and formatters configured. my_script.py configures handlers and formatters using fileConfig and a config file that lives alongside my_script.py:
import os
import logging
import logging.config
from my_mod.module import my_command
logging.config.fileConfig('{0}/logging.cfg'.format(
os.path.dirname(os.path.realpath(__file__))))
LOGGER = logging.getLogger(__name__)
LOGGER.debug("This is a log message from script.py")
my_command()
From what I can tell, my config file appears to be set up correctly...
[loggers]
keys=root,script
[handlers]
keys=consoleHandler
[formatters]
keys=simpleFormatter
[logger_root]
level=DEBUG
handlers=consoleHandler
[logger_script]
level=DEBUG
handlers=consoleHandler
qualname=script
propagate=0
[handler_consoleHandler]
class=StreamHandler
level=DEBUG
formatter=simpleFormatter
args=(sys.stdout,)
[formatter_simpleFormatter]
format=%(asctime)s [%(levelname)s] %(name)s: %(message)s
datefmt=
...but when I run my_script.py I get only the log line from my_script.py, and not the one from my_mod.my_command. I know that my_command is working, though, because the print statement in my_command after the debug log statement successfully prints to the console:
20:27 $ python script.py
2015-06-15 20:27:54,488 [DEBUG] __main__: This is a log message from script.py
This is a print statement from module.py
What am I doing wrong?
NOTE: The example shows using debug, but even when I keep logging.cfg specifying level=DEBUG (I also tried level=NOTSET) for root logger and call LOGGER.info(message) in my_command, nothing is logged to the console.

A potential problem is that you are importing the module before you set up the logger configuration. That way, the module requests a logger before the logging is setup.
Looking a fileConfig()'s documentation, the reason subsequent logging to the pre-obtained loggers fails is the default value for its disable_existing_loggers argument:
logging.config.fileConfig(fname, defaults=None, disable_existing_loggers=True)
If you change your code to
logging.config.fileConfig(
'{0}/logging.cfg'.format(os.path.dirname(os.path.realpath(__file__))),
disable_existing_loggers=False
)
the problem should go away.
Note that existing loggers are only disabled when they are not explicitly named in the configuration file. For example:
import logging
import logging.config
lFooBefore = logging.getLogger('foo')
lScriptBefore = logging.getLogger('script')
logging.config.fileConfig('logger.ini')
lFooBefore.debug('Does not log')
lScriptBefore.debug('Does log')
logging.getLogger('foo').debug('Does also not log')
logging.getLogger('bar').debug('Does log')
No idea why the default value for disable_existing_loggers is the way it is ...

No handlers found for logger __main__

I set up logging throughout my python package using a logconfig.ini file.
[loggers]
keys=extracts,root
[formatters]
keys=simple,detailed
[handlers]
keys=file_handler
[formatter_simple]
format=%(module)s - %(levelname)s - %(message)s
datefmt=%Y-%m-%d %H:%M:%S
[formatter_detailed]
format=%(asctime)s %(name)s:%(lineno)s %(levelname)s %(message)s
datefmt=%Y-%m-%d %H:%M:%S
[handler_file_handler]
class=logging.handlers.RotatingFileHandler
level=DEBUG
formatter=detailed
args=('/ebs/logs/foo.log', 'a', 100000000, 3)
[logger_extracts]
level=DEBUG
handlers=file_handler
propagate=1
qualname=extracts
[logger_root]
level=NOTSET
handlers=
But whenever I run my application, I get the following warning message in prompt,
No handlers found for logger __main__
How can I fix this?

You have to call logging.basicConfig() first:
Logging HOWTO
The call to basicConfig() should come before any calls to debug(),
info() etc. As it’s intended as a one-off simple configuration
facility, only the first call will actually do anything: subsequent
calls are effectively no-ops.
Or all logging.info('Starting logger for...') which will call logging.basicConfig() automatically. So something like:
import logging
logging.info('Starting logger for...') # or call logging.basicConfig()
LOG = logging.getLogger(name)
The module author's reason for this behavior is here

I found my error.
It turns out the the root logger is used for main.
I just need to attach a handler to the root logger as so,
[logger_root]
level=NOTSET
handlers=file_handler

Python sockethandler logging not working for some modules

I have an odd problem. I am using logging.config to setup my logger to use the sockethandler. Everything works fine except 2/10+ modules don't seem to be logging anything. By fine, I mean I see output on my log server for everything except for the two. It's so odd that it works for some but not others. I'm initializing the logger for every module with the following lines:
import logging
import logging.config
logging.config.fileConfig(config.main_log_conf)
logger = logging.getLogger("CAKE")
I thought the problem was might've been conflicting logger names, thus the CAKE above, but that didn't work.
Below is the conf file I'm using.
[loggers]
keys=root
[handlers]
keys=socketHandler
[formatters]
keys=simpleFormatter
[logger_root]
handlers=socketHandler
level=DEBUG
[handler_socketHandler]
class=handlers.SocketHandler
level=DEBUG
formatter=simpleFormatter
args=('localhost', handlers.DEFAULT_TCP_LOGGING_PORT)
[formatter_simpleFormatter]
format=%(asctime)s - %(name)s - %(levelname)s - %(message)s
datefmt=

Don't call fileConfig() from every module. There should generally be just one call to set up the logging configuration in a program (process), called from the if __name__ == '__main__' clause. Also note that you should (in general) pass disable_existing_loggers=False to fileConfig - the default is True only for backward compatibility reasons. See here for more information.

How to set more than one handler through config file in python?

I know you can set more than one logger handler with different level by calling addHandler in Python.
However, I need to set two handlers through a config file, so that by calling
logging.config.fileConfig('log.conf')
logger = logging.getLogger('my_logger')
I get a logger that have two handlers for different level.
How should I write the log.conf file?

Here's a basic example with the root logger having two handlers:
[loggers]
keys=root
[handlers]
keys=hand01,hand02
[formatters]
keys=
[logger_root]
level=DEBUG
handlers=hand01,hand02
[handler_hand01]
class=StreamHandler
args=()
[handler_hand02]
class=StreamHandler
args=()

Logging in Python with Config File - Using handlers defined in file through code

I am using logging module of python. How can I access the handlers defined in config file from the code. As an example, I have a logger defined and two handlers - one for screen and other for file. I want to use appropriate handler based on user preference (whether they want to log on screen or to a file). How can I dynamically add and remove handlers defined in config file from loggers defined in config file?
[loggers]
keys=root,netmap
[handlers]
keys=fil,screen
[logger_root]
level=NOTSET
handlers=
[logger_netmap]
level=INFO
handlers=fil,screen
qualname=netmap
[formatters]
keys = simple
[formatter_simple]
format=%(asctime)s - %(name)s - %(levelname)s - %(message)s
datefmt=
[handler_fil]
class=handlers.RotatingFileHandler
args=('file.log','a','maxBytes=10000','backupCount=5')
formatter=simple
[handler_screen]
class=StreamHandler
args = (sys.stdout,)
formatter=simple
Depending on whether user runs the program with -v or not I need to use one of File or Screen Handler. How can I add or delete fil or screen handlers from netmap logger?

Instead of having to dynamically change the config file, just use Logger.addHandler(handler).
fileHandler = logging.handlers.RotatingFileHandler('file.log', mode='a', maxBytes=10000, backupCount=5)
logger = logging.getLogger('netmap')
if LOG_TO_FILE:
logger.addHandler(fileHandler)
Then to load in formatting from perhaps the same file;
import ConfigParser
configParser = ConfigParser.ConfigParser()
config.read('config.ini')
format = config.get('formatter_simple', 'format')
fileHandler.setFormatter(format)

From the logging module's documentation it looks like logging objects have these two methods:
Logger.addHandler(hdlr)
Adds the specified handler hdlr to this logger.
Logger.removeHandler(hdlr)
Removes the specified handler hdlr from this logger.
So it seems like you ought be able to use them to add or change the netmap logger handler to be whatever you want based on what's in the config file.

Ok So I found an elegant way through a gud soul on the python google group. It works like charm. Here's the code
import logging
import getopt
import sys
import logging.config
def stop(m,handl):
consoleHand=[h for h in m.handlers if h.__class__ is handl]
print consoleHand
if consoleHand:
h1=consoleHand[0]
h1.filter=lambda x: False
logging.config.fileConfig('log.conf')
my_log = logging.getLogger('netmap')
args=sys.argv[1:]
opt,arg = getopt.gnu_getopt(args,'v')
l=''
for o,a in opt:
if o=='-v':
l='verbose'
stop(my_log,logging.handlers.RotatingFileHandler)
if not l:
stop(my_log,logging.StreamHandler)
my_log.debug('Starting.....')
my_log.warning('Unstable')
my_log.error('FIles Missing')
my_log.critical('BlowOut!!')
The config file is still the same. I can now easily activate or deactivate handlers.