Logging¶

Logging errors¶

The preferred way to log errors is by using the Flask application logger:

>>> with app.app_context():
...     try:
...         raise Exception("This is an exception")
...     except Exception:
...         current_app.logger.exception("My message")

logger.exception() will automatically include the exception stacktrace in the log record, which each log handler may decide to include or not.

You may also manually include exception information in the logger using the exc_info keyword argument:

>>> import sys
>>> with app.app_context():
...     try:
...         raise Exception("This is an exception")
...     except Exception:
...         current_app.logger.critical("My message", exc_info=1)

Naturally, other log levels may also be used:

>>> app.logger.info("This is an info message")
>>> app.logger.warning("This is a warning message")
>>> app.logger.debug("This is a debug message")

Log handlers¶

Log messages written to the Flask application logger can be handled by many different backends, which is configurable by the an administrator. By default Invenio ships with following log handlers:

• invenio_ext.logging.backends.fs - Rotating file system handler.
• invenio_ext.logging.backends.legacy - Error email reporting and logging to database. Default logging behaviour of Invenio 1.x.
• invenio_ext.logging.backends.sentry - Logging to Sentry service (see https://pypi.python.org/pypi/sentry and https://getsentry.com/)

Installing one or more of the logging backends is a simple as including them in your configuration variable EXTENSIONS:

EXTENSIONS = [
    # ...
    'invenio_ext.logging',
    'invenio_ext.logging.backends.fs',
    'invenio_ext.logging.backends.legacy',
    'invenio_ext.logging.backends.sentry',
    # ...
]

Note that each backend may require additional configuration. Please see Backends for specific details.

Additionally if you plan to write your own backend, you may wish to consult Python’s logging documentation for how to create handlers, formatters and filters: https://docs.python.org/2/library/logging.html

Legacy handling of errors¶

Invenio 1.x used a method register_exception to log errors. This method may still be used, but may be deprecated in the future:

>>> from invenio_ext.logging import register_exception
>>> with app.app_context():
...     try:
...         raise Exception("This is an exception")
...     except Exception as e:
...         register_exception()

The method register_exception is in fact just a small wrapper around the application logger.

Error handling do’s and don’ts¶

Always use ``except Exception:`` (or preferably more specific exceptions) over ``except:``, unless you explicitly want to catch the following built-in exceptions (SystemExit, KeyboardInterrupt, GeneratorExit).

See https://docs.python.org/2/library/exceptions.html#exception-hierarchy

Reraise. To gracefully handle errors, you may often catch exceptions to perform some cleanup or e.g. convert a low-level library exception into a more high-level application exception. This may however often discard the initial exception and its traceback, making it hard to track down the root cause. To preserve the traceback, simply reraise the caught exception using a raise with no arguments:

>>> with app.app_context():
...     try:
...         try:
...             0 / 0
...         except ZeroDivisionError as e:
...             # Do clean-up
...             raise
...     except Exception as e:
...         current_app.logger.exception("Something bad happened")

If you like to convert the exception, it can be done like this:

>>> import six
>>> class AppError(Exception):
...     pass
>>> with app.app_context():
...     try:
...         try:
...             0 / 0
...         except ZeroDivisionError as e:
...             six.reraise(MyAppError, "Custom message", sys.exc_info()[2])
...     except AppError as e:
...         current_app.logger.exception("Something bad happened")

Backends¶