Logging

A quick logging primer

Django uses Python’s builtin logging module to perform system logging. The usage of this module is discussed in detail in Python’s own documentation. However, if you’ve never used Python’s logging framework (or even if you have), here’s a quick primer.

The cast of players

A Python logging configuration consists of four parts:

Loggers

A logger is the entry point into the logging system. Each logger is a named bucket to which messages can be written for processing.

A logger is configured to have a log level. This log level describes the severity of the messages that the logger will handle. Python defines the following log levels:

  • DEBUG: Low level system information for debugging purposes
  • INFO: General system information
  • WARNING: Information describing a minor problem that has occurred.
  • ERROR: Information describing a major problem that has occurred.
  • CRITICAL: Information describing a critical problem that has occurred.

Each message that is written to the logger is a Log Record. Each log record also has a log level indicating the severity of that specific message. A log record can also contain useful metadata that describes the event that is being logged. This can include details such as a stack trace or an error code.

When a message is given to the logger, the log level of the message is compared to the log level of the logger. If the log level of the message meets or exceeds the log level of the logger itself, the message will undergo further processing. If it doesn’t, the message will be ignored.

Once a logger has determined that a message needs to be processed, it is passed to a Handler.

Handlers

The handler is the engine that determines what happens to each message in a logger. It describes a particular logging behavior, such as writing a message to the screen, to a file, or to a network socket.

Like loggers, handlers also have a log level. If the log level of a log record doesn’t meet or exceed the level of the handler, the handler will ignore the message.

A logger can have multiple handlers, and each handler can have a different log level. In this way, it is possible to provide different forms of notification depending on the importance of a message. For example, you could install one handler that forwards ERROR and CRITICAL messages to a paging service, while a second handler logs all messages (including ERROR and CRITICAL messages) to a file for later analysis.

Filters

A filter is used to provide additional control over which log records are passed from logger to handler.

By default, any log message that meets log level requirements will be handled. However, by installing a filter, you can place additional criteria on the logging process. For example, you could install a filter that only allows ERROR messages from a particular source to be emitted.

Filters can also be used to modify the logging record prior to being emitted. For example, you could write a filter that downgrades ERROR log records to WARNING records if a particular set of criteria are met.

Filters can be installed on loggers or on handlers; multiple filters can be used in a chain to perform multiple filtering actions.

Formatters

Ultimately, a log record needs to be rendered as text. Formatters describe the exact format of that text. A formatter usually consists of a Python formatting string; however, you can also write custom formatters to implement specific formatting behavior.

Using logging

Once you have configured your loggers, handlers, filters and formatters, you need to place logging calls into your code. Using the logging framework is very simple. Here’s an example:

# import the logging library
import logging

# Get an instance of a logger
logger = logging.getLogger(__name__)

def my_view(request, arg1, arg):
    ...
    if bad_mojo:
        # Log an error message
        logger.error('Something went wrong!')

And that’s it! Every time the bad_mojo condition is activated, an error log record will be written.

Naming loggers

The call to logging.getLogger() obtains (creating, if necessary) an instance of a logger. The logger instance is identified by a name. This name is used to identify the logger for configuration purposes.

By convention, the logger name is usually __name__, the name of the python module that contains the logger. This allows you to filter and handle logging calls on a per-module basis. However, if you have some other way of organizing your logging messages, you can provide any dot-separated name to identify your logger:

# Get an instance of a specific named logger
logger = logging.getLogger('project.interesting.stuff')

The dotted paths of logger names define a hierarchy. The project.interesting logger is considered to be a parent of the project.interesting.stuff logger; the project logger is a parent of the project.interesting logger.

Why is the hierarchy important? Well, because loggers can be set to propagate their logging calls to their parents. In this way, you can define a single set of handlers at the root of a logger tree, and capture all logging calls in the subtree of loggers. A logging handler defined in the project namespace will catch all logging messages issued on the project.interesting and project.interesting.stuff loggers.

This propagation can be controlled on a per-logger basis. If you don’t want a particular logger to propagate to it’s parents, you can turn off this behavior.

Making logging calls

The logger instance contains an entry method for each of the default log levels:

  • logger.critical()
  • logger.error()
  • logger.warning()
  • logger.info()
  • logger.debug()

There are two other logging calls available:

  • logger.log(): Manually emits a logging message with a specific log level.
  • logger.exception(): Creates an ERROR level logging message wrapping the current exception stack frame.

Configuring logging

Of course, it isn’t enough to just put logging calls into your code. You also need to configure the loggers, handlers, filters and formatters to ensure that logging output is output in a useful way.

Python’s logging library provides several techniques to configure logging, ranging from a programmatic interface to configuration files. By default, Django uses the dictConfig format.

Note

logging.dictConfig is a builtin library in Python 2.7. In order to make this library available for users of earlier Python versions, Django includes a copy as part of django.utils.log. If you have Python 2.7, the system native library will be used; if you have Python 2.6 or earlier, Django’s copy will be used.

In order to configure logging, you use LOGGING to define a dictionary of logging settings. These settings describes the loggers, handlers, filters and formatters that you want in your logging setup, and the log levels and other properties that you want those components to have.

Logging is configured immediately after settings have been loaded. Since the loading of settings is one of the first things that Django does, you can be certain that loggers are always ready for use in your project code.

An example

The full documentation for dictConfig format is the best source of information about logging configuration dictionaries. However, to give you a taste of what is possible, here is an example of a fairly complex logging setup, configured using logging.dictConfig():

LOGGING = {
    'version': 1,
    'disable_existing_loggers': True,
    'formatters': {
        'verbose': {
            'format': '%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d %(message)s'
        },
        'simple': {
            'format': '%(levelname)s %(message)s'
        },
    },
    'filters': {
        'special': {
            '()': 'project.logging.SpecialFilter',
            'foo': 'bar',
        }
    },
    'handlers': {
        'null': {
            'level':'DEBUG',
            'class':'django.utils.log.NullHandler',
        },
        'console':{
            'level':'DEBUG',
            'class':'logging.StreamHandler',
            'formatter': 'simple'
        },
        'mail_admins': {
            'level': 'ERROR',
            'class': 'django.utils.log.AdminEmailHandler',
            'filters': ['special']
        }
    },
    'loggers': {
        'django': {
            'handlers':['null'],
            'propagate': True,
            'level':'INFO',
        },
        'django.request': {
            'handlers': ['mail_admins'],
            'level': 'ERROR',
            'propagate': False,
        },
        'myproject.custom': {
            'handlers': ['console', 'mail_admins'],
            'level': 'INFO',
            'filters': ['special']
        }
    }
}

This logging configuration does the following things:

  • Identifies the configuration as being in ‘dictConfig version 1’ format. At present, this is the only dictConfig format version.

  • Disables all existing logging configurations.

  • Defines two formatters:

    • simple, that just outputs the log level name (e.g., DEBUG) and the log message.

      The format string is a normal Python formatting string describing the details that are to be output on each logging line. The full list of detail that can be output can be found in the formatter documentation.

    • verbose, that outputs the log level name, the log message, plus the time, process, thread and module that generate the log message.

  • Defines one filter – project.logging.SpecialFilter, using the alias special. If this filter required additional arguments at time of construction, they can be provided as additional keys in the filter configuration dictionary. In this case, the argument foo will be given a value of bar when instantiating the SpecialFilter.

  • Defines three handlers:

    • null, a NullHandler, which will pass any DEBUG (or higher) message to /dev/null.
    • console, a StreamHandler, which will print any DEBUG (or higher) message to stderr. This handler uses the simple output format.
    • mail_admins, an AdminEmailHandler, which will email any ERROR (or higher) message to the site admins. This handler uses the special filter.
  • Configures three loggers:

    • django, which passes all messages at INFO or higher to the null handler.
    • django.request, which passes all ERROR messages to the mail_admins handler. In addition, this logger is marked to not propagate messages. This means that log messages written to django.request will not be handled by the django logger.
    • myproject.custom, which passes all messages at INFO or higher that also pass the special filter to two handlers – the console, and mail_admins. This means that all INFO level messages (or higher) will be printed to the console; ERROR and CRITICAL messages will also be output via email.

Custom handlers and circular imports

If your settings.py specifies a custom handler class and the file defining that class also imports settings.py a circular import will occur.

For example, if settings.py contains the following config for LOGGING:

LOGGING = {
  'version': 1,
  'handlers': {
    'custom_handler': {
      'level': 'INFO',
      'class': 'myproject.logconfig.MyHandler',
    }
  }
}

and myproject/logconfig.py has the following line before the MyHandler definition:

from django.conf import settings

then the dictconfig module will raise an exception like the following:

ValueError: Unable to configure handler 'custom_handler':
Unable to configure handler 'custom_handler':
'module' object has no attribute 'logconfig'

Custom logging configuration

If you don’t want to use Python’s dictConfig format to configure your logger, you can specify your own configuration scheme.

The LOGGING_CONFIG setting defines the callable that will be used to configure Django’s loggers. By default, it points at Python’s logging.dictConfig() method. However, if you want to use a different configuration process, you can use any other callable that takes a single argument. The contents of LOGGING will be provided as the value of that argument when logging is configured.

Disabling logging configuration

If you don’t want to configure logging at all (or you want to manually configure logging using your own approach), you can set LOGGING_CONFIG to None. This will disable the configuration process.

Note

Setting LOGGING_CONFIG to None only means that the configuration process is disabled, not logging itself. If you disable the configuration process, Django will still make logging calls, falling back to whatever default logging behavior is defined.

Django’s logging extensions

Django provides a number of utilities to handle the unique requirements of logging in Web server environment.

Loggers

Django provides three built-in loggers.

django

django is the catch-all logger. No messages are posted directly to this logger.

django.request

Log messages related to the handling of requests. 5XX responses are raised as ERROR messages; 4XX responses are raised as WARNING messages.

Messages to this logger have the following extra context:

  • status_code: The HTTP response code associated with the request.
  • request: The request object that generated the logging message.

django.db.backends

Messages relating to the interaction of code with the database. For example, every SQL statement executed by a request is logged at the DEBUG level to this logger.

Messages to this logger have the following extra context:

  • duration: The time taken to execute the SQL statement.
  • sql: The SQL statement that was executed.
  • params: The parameters that were used in the SQL call.

For performance reasons, SQL logging is only enabled when settings.DEBUG is set to True, regardless of the logging level or handlers that are installed.

Handlers

Django provides one log handler in addition to those provided by the Python logging module.

class AdminEmailHandler([include_html=False])

This handler sends an email to the site admins for each log message it receives.

If the log record contains a request attribute, the full details of the request will be included in the email.

If the log record contains stack trace information, that stack trace will be included in the email.

The include_html argument of AdminEmailHandler is used to control whether the traceback email includes an HTML attachment containing the full content of the debug Web page that would have been produced if DEBUG were True. To set this value in your configuration, include it in the handler definition for django.utils.log.AdminEmailHandler, like this:

'handlers': {
    'mail_admins': {
        'level': 'ERROR',
        'class': 'django.utils.log.AdminEmailHandler',
        'include_html': True,
    }
},

Note that this HTML version of the email contains a full traceback, with names and values of local variables at each level of the stack, plus the values of your Django settings. This information is potentially very sensitive, and you may not want to send it over email. Consider using something such as django-sentry to get the best of both worlds – the rich information of full tracebacks plus the security of not sending the information over email. You may also explicitly designate certain sensitive information to be filtered out of error reports – learn more on Filtering error reports.

Filters

Django provides two log filters in addition to those provided by the Python logging module.

class CallbackFilter(callback)

This filter accepts a callback function (which should accept a single argument, the record to be logged), and calls it for each record that passes through the filter. Handling of that record will not proceed if the callback returns False.

class RequireDebugFalse

This filter will only pass on records when settings.DEBUG is False.

This filter is used as follows in the default LOGGING configuration to ensure that the AdminEmailHandler only sends error emails to admins when DEBUG is False:

'filters': {
     'require_debug_false': {
         '()': 'django.utils.log.RequireDebugFalse',
     }
 },
 'handlers': {
     'mail_admins': {
         'level': 'ERROR',
         'filters': ['require_debug_false'],
         'class': 'django.utils.log.AdminEmailHandler'
     }
 },