log messages appearing twice with Python Logging

The handler is added each time you call from outside. Try Removeing the Handler after you finish your job:

self.logger.removeHandler(ch)

I’m a python newbie, but this seemed to work for me (Python 2.7)

while logger.handlers:
     logger.handlers.pop()

If you are seeing this problem and you’re not adding the handler twice then see abarnert’s answer here

From the docs:

Note: If you attach a handler to a logger and one or more of its
ancestors, it may emit the same record multiple times. In general, you
should not need to attach a handler to more than one logger – if you
just attach it to the appropriate logger which is highest in the
logger hierarchy, then it will see all events logged by all descendant
loggers, provided that their propagate setting is left set to True. A
common scenario is to attach handlers only to the root logger, and to
let propagation take care of the rest.

So, if you want a custom handler on "test", and you don’t want its messages also going to the root handler, the answer is simple: turn off its propagate flag:

logger.propagate = False

A call to logging.debug() calls logging.basicConfig() if there are no root handlers installed. That was happening for me in a test framework where I couldn’t control the order that test cases fired. My initialization code was installing the second one. The default uses logging.BASIC_FORMAT that I didn’t want.

It seems that if you output something to the logger (accidentally) then configure it, it is too late. For example, in my code I had

logging.warning("look out)"

...
ch = logging.StreamHandler(sys.stdout)
root = logging.getLogger()
root.addHandler(ch)

root.info("hello")

I would get something like (ignoring the format options)

look out
hello
hello

and everything was written to stdout twice. I believe this is because the first call to logging.warning creates a new handler automatically, and then I explicitly added another handler. The problem went away when I removed the accidental first logging.warning call.

In my case I’d to set logger.propagate = False to prevent double printing.

In below code if you remove logger.propagate = False then you will see double printing.

import logging
from typing import Optional

_logger: Optional[logging.Logger] = None

def get_logger() -> logging.Logger:
    global _logger
    if _logger is None:
        raise RuntimeError('get_logger call made before logger was setup!')
    return _logger

def set_logger(name:str, level=logging.DEBUG) -> None:
    global _logger
    if _logger is not None:
        raise RuntimeError('_logger is already setup!')
    _logger = logging.getLogger(name)
    _logger.handlers.clear()
    _logger.setLevel(level)
    ch = logging.StreamHandler()
    ch.setLevel(level)
    # warnings.filterwarnings("ignore", "(Possibly )?corrupt EXIF data", UserWarning)
    ch.setFormatter(_get_formatter())
    _logger.addHandler(ch)
    _logger.propagate = False # otherwise root logger prints things again


def _get_formatter() -> logging.Formatter:
    return logging.Formatter(
        '[%(asctime)s] [%(name)s] [%(levelname)s] %(message)s')

I was getting a strange situation where console logs were doubled but my file logs were not. After a ton of digging I figured it out.

Please be aware that third party packages can register loggers. This is something to watch out for (and in some cases can’t be prevented). In many cases third party code checks to see if there are any existing root logger handlers; and if there isn’t–they register a new console handler.

My solution to this was to register my console logger at the root level:

rootLogger = logging.getLogger()  # note no text passed in--that's how we grab the root logger
if not rootLogger.handlers:
        ch = logging.StreamHandler()
        ch.setLevel(logging.INFO)
        ch.setFormatter(logging.Formatter('%(process)s|%(levelname)s] %(message)s'))
        rootLogger.addHandler(ch)

I was struggling with the same issue in the context of multiple processes. (For the code see the docs which I was following almost verbatim.) Namely, all log messages originating from any of the child processes got duplicated.

My mistake was to call worker_configurer(),

def worker_configurer(logging_queue):
    queue_handler = logging.handlers.QueueHandler(logging_queue)
    root = logging.getLogger()
    root.addHandler(queue_handler)
    root.setLevel(level)

both in the child processes and also in the main process (since I wanted the main process to log stuff, too). The reason this led to trouble (on my Linux machine) is that on Linux the child processes got started through forking and therefore inherited the existing log handlers from the main process. That is, on Linux the QueueHandler got registered twice.

Now, preventing the QueueHandler from getting registered twice in the worker_configurer() function is not as trivial as it seems:

• Logger objects like the root logger root have a handlers property but it is undocumented.

• In my experience, testing whether any([handler is queue_handler for handler in root.handlers]) (identity) or any([handler == queue_handler for handler in root.handlers]) (equality) fails after forking, even if root.handlers seemingly contains the same QueueHandler. (Obviously, the previous two expressions can be abbreviated by queue_handler in root.handlers, since the in operator checks for both identity and equality in the case of lists.)

• The root logger gets modified by packages like pytest, so root.handlers and root.hasHandlers() are not very reliable to begin with. (They are global state, after all.)

The clean solution, therefore, is to replace forking with spawning to prevent these kinds of multiprocessing bugs right from the start (provided you can live with the additional memory footprint, of course). Or to use an alternative to the logging package that doesn’t rely on global state and instead requires you to do proper dependency injection but I’m digressing… 🙂

With that being said, I ended up going for a rather trivial check:

def worker_configurer(logging_queue):
    queue_handler = logging.handlers.QueueHandler(logging_queue)
    root = logging.getLogger()

    for handler in root.handlers:
        if isinstance(handler, logging.handlers.QueueHandler):
            return

    root.addHandler(queue_handler)
    root.setLevel(level)

Obviously, this will have nasty side effects the second I decide to register a second queue handler somewhere else.

If you are using any config for logging, For instance log.conf

In .conf file you can do it by adding this line in the [logger_myLogger] section: propagate=0

[logger_myLogger]
level=DEBUG
handlers=validate,consoleHandler
qualname=VALIDATOR
propagate=0

This can also happen if you are trying to create a logging object from the parent file. For e.g.
This is the main application file test.py

import logging

# create logger
logger = logging.getLogger('simple_example')
logger.setLevel(logging.DEBUG)

# create console handler and set level to debug
ch = logging.StreamHandler()
ch.setLevel(logging.DEBUG)

# create formatter
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')

# add formatter to ch
ch.setFormatter(formatter)

# add ch to logger
logger.addHandler(ch)

def my_code():
# 'application' code
    logger.debug('debug message')
    logger.info('info message')
    logger.warning('warn message')
    logger.error('error message')
    logger.critical('critical message')

And below is the parent file main.py

import test

test.my_code()

The output of this will print only once

2021-09-26 11:10:20,514 - simple_example - DEBUG - debug message
2021-09-26 11:10:20,514 - simple_example - INFO - info message
2021-09-26 11:10:20,514 - simple_example - WARNING - warn message
2021-09-26 11:10:20,514 - simple_example - ERROR - error message
2021-09-26 11:10:20,514 - simple_example - CRITICAL - critical message

But if we had a parent logging object, then it will be printed twice. For e.g. if this is the parent file

import test
import logging
logging.basicConfig(level=logging.DEBUG,format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')


test.my_code()

The the output will be

2021-09-26 11:16:28,679 - simple_example - DEBUG - debug message
2021-09-26 11:16:28,679 - simple_example - DEBUG - debug message
2021-09-26 11:16:28,679 - simple_example - INFO - info message
2021-09-26 11:16:28,679 - simple_example - INFO - info message
2021-09-26 11:16:28,679 - simple_example - WARNING - warn message
2021-09-26 11:16:28,679 - simple_example - WARNING - warn message
2021-09-26 11:16:28,679 - simple_example - ERROR - error message
2021-09-26 11:16:28,679 - simple_example - ERROR - error message
2021-09-26 11:16:28,679 - simple_example - CRITICAL - critical message
2021-09-26 11:16:28,679 - simple_example - CRITICAL - critical message

From the docs:

"Loggers have the following attributes and methods. Note that Loggers should NEVER be instantiated directly, but always through the module-level function logging.getLogger(name).
Multiple calls to getLogger() with the same name will always return a reference to the same Logger object".

Make sure you don’t initialise your loggers with the same name
I advise you to initialise the logger with __name__ as name param i.e:

import logging
logger = logging.getLogger(__name__)

NOTE:
even if you init a loggers from other modules with same name, you will still get the same logger, therefore calling i.e logger.info('somthing') will log as many times as you initiated the logger class.

If you use the standard construction logger = logging.getLogger('mymodule') and then accidentally mistype loggger as logging i.e.

logger = logging.getLogger('mymodule')

# configure your handlers

logger.info("my info message")  # won't make duplicate 
logging.info("my info message")  # will make duplicate logs

then this will cause duplicate messages to come up because the call to logging creates a new logger.

I had the same issue. In my case, it was not due to handlers or duplicate initial configuration but a stupid typo. In main.py I was using a logger object but in my_tool.py I was directly calling to the logging module by mistake, hence after invoking functions from my_tool module everything was messed up and the messages appeared duplicated.

This was the code:

main.py

import logging
import my_tool

logger_name = "cli"
logger = logging.getLogger(logger_name)

logger.info("potato")
logger.debug("potato)
my_tool.function()
logger.info("tomato")

my_tool.py

import logging
logger_name = "cli"
logger = logging.getLogger(logger_name)
# some code
logging.info("carrot")

and the result

terminal

>> potato
>> potato
>> carrot
>> tomato
>> tomato

Adding to others’ useful responses…

In case you are NOT accidentally configuring more than one handler on your logger:

When your logger has ancestors(root logger is always one) and they have their own handlers, they will also output when your logger outputs (by default), which will create duplicates. You have two options:

• Don’t propagate your log event to your ancestors by setting:

my_logger.propagate = False

• If you only have one ancestor(root logger), you could directly configure their handler instead. For example:

# directly change the formatting of root's handler
root_logger = logging.getLogger()
roots_handler = root_logger.handlers[0]
roots_handler.setFormatter(logging.Formatter(': %(message)s')) # change format

my_logger = logging.getLogger('my_logger') # my_logger will use new formatting

I tried many things and finally came up with my custom logger. You just need to copy and paste it in your code and it will do the work you required.

Why to use?

1. No repeated logs.
2. Separate customisation for info logs – [ Not contains path and unnecessary details ]
3. Different colours to different log levels.
4. You will get all the different levels of logs – [ info, warning, error, critical, debug ,… ].

import logging


class CustomFormatter(logging.Formatter):
    def __init__(self) -> None:
        super().__init__()

    def format(self, record):
        if record.levelno == logging.INFO:
            self._style._fmt = "%(levelname)-5s  %(asctime)-3s %(message)s"
        else:
            color = {
                logging.WARNING: 33,
                logging.ERROR: 31,
                logging.FATAL: 31,
                # logging.DEBUG: 36
            }.get(record.levelno, 0)
            self._style._fmt = f'33[{color}m%(levelname)-5s %(asctime)s file/"%(pathname)10s", line %(lineno)d in "%(funcName)s" -> "%(message)s" 33[0m'
        return super().format(record)


logger = logging.getLogger("my-logger")
if not logger.handlers:
    handler = logging.StreamHandler()
    handler.setFormatter(CustomFormatter())
    logger.addHandler(handler)
    logger.setLevel(logging.DEBUG)
    logger.propagate = False
    # to disable any level of logs; eg - disabled debug level logs.
    logging.disable(logging.DEBUG)

What you will get [ *it will be with different colours in you colsole/log file ]

logger.info("This is info log")

INFO 2023-03-24 20:11:51,230 This is info log

logger.warning("This is warning log")

WARNING 2023-03-24 20:11:51,230 file/"/home/xyz/abc", line 25 in "my_function" -> "This is warning log"

logger.error("This is error log")

ERROR 2023-03-24 20:11:51,230 file/"/home/xyz/abc", line 26 in "my_function" -> "This is error log"

logger.critical("This is critical log")

CRITICAL 2023-03-24 20:11:51,231 file/"/home/xyz/abc", line 27 in "my_function" -> "This is critical log"

logger.debug("This is debug log")

DEBUG 2023-03-24 20:11:51,231 file/"/home/xyz/abc", line 28 in "my_function" -> "This is debug log"

Note – I have disabled the debug level log in the above code, to use comment/remove the lines.