This repository contains the code of reconplogger, a python package intended to ease the standardization of logging within omni:us. The main design decision of reconplogger is to allow total freedom to reconfigure loggers without hard coding anything.
The package contains essentially the following things:
- A default logging configuration.
- A function for loading logging configuration for regular python code.
- A function for loading logging configuration for flask-based microservices.
- An inheritable class to add a logger property.
- A context manager to set and get the correlation id.
- Lower level functions for:
- Loading logging configuration from any of: config file, environment variable, or default.
- Replacing the handlers of an existing Logger object.
- Function to add a file handler to a logger.
There are two main use cases reconplogger targets. One is for logging in regular generic python code and the second one is logging in microservices. See the two standardizing sections below for a detailed explanation of the two use cases.
The first step to use reconplogger is adding it as a requirement in the
respective package where it will be used. This means adding it in the file
setup.cfg as an item in install_requires
or in an
extras_require
depending on whether reconplogger is intended to be a
core or an optional requirement.
Note: It is highly discouraged to develop packages in which requirements are added directly to setup.py or to have an ambiguous requirements.txt file. See the setup.cfg file in the reconplogger source code for reference.
A feature that reconplogger provides is the possibility of externally setting the logging configuration without having to change code or implement any parsing of configuration. However, if a logging configuration is not given externally, reconplogger provides a default configuration.
The default configuration defines three handlers, two of which are stream
handlers and are set to DEBUG log level. The first handler called
plain_handler
uses a simple plain text formatter, and the second handler
called json_handler
as the name suggests outputs in json format, using
the logmatic JsonFormatter class.
The third handler called null_handler
is useful to disable all logging.
For each handler the default configuration defines a corresponding logger:
plain_logger
, json_logger
and null_logger
.
One objective of reconplogger is to ease the use of logging and standardize the way it is done across all omni:us python code. The use of reconplogger comes down to calling one function to get the logger object. For regular python code (i.e. not a microservice) the function to use is :func:`reconplogger.logger_setup`.
The following code snippet illustrates the use:
import reconplogger
# Default plain logger
logger = reconplogger.logger_setup()
logger.info('My log message')
# Json logger and custom prefix
logger = reconplogger.logger_setup('json_logger', env_prefix='MYAPP')
logger.info('My log message in json format')
This function gives you the ability to set the default logger to use
(logger_name
argument whose default value is plain_logger
) and
optionally provide a logging config
and/or a logging level
that
overrides the level in the config.
All of these values can be overridden via environment variables whose names are
prefixed by the value of the env_prefix
argument. The environment
variables supported are: {env_prefix}_CFG
, {env_prefix}_NAME
and
{env_prefix}_LEVEL
. Note that the environment variable names are not
required to be prefixed by the default env_prefix='LOGGER'
. The prefix
can be chosen by the user for each particular application.
For functions or classes that receive logger object as an argument, it might be
desired to set a non-logging default so that it can be called without specifying
one. For this reconplogger defines null logger
that could be used as
follows:
from reconplogger import null_logger
...
def my_func(arg1, arg2, logger=null_logger):
...
The most important objective of reconplogger is to allow standardization of a structured logging format for all microservices developed. Thus, the logging from all microservices should be configured like explained here. The use is analogous to the previous case, but using the function :func:`reconplogger.flask_app_logger_setup` instead, and giving as first argument the flask app object.
Additional to the previous case, this function:
- Replaces the flask app and werkzeug loggers to use a reconplogger configured one.
- Add to the logs the correlation_id
- Add before and after request functions to log the request details when the request is processed
- Patch the requests library forwarding the correlation id in any call to other microservices
What is the correlation ID? In a system build with microservices we need a way to correlate logs coming from different microservices to the same "external" call. For example when a user of our system do a call to the MicroserviceA this could need to retrieve some information from the MicroserviceB, if there is an error and we want to check the logs of the MicroserviceB related to the user call we don't have a way to correlate them, to solve this we use the correlation id! It is a unique string that is passed in the headers of the REST calls and will be forwarded automatically when we do calls with the library requests. All of this is taken care in the background by this library. If the correlation id its not present in the request headers, it will not be generated. It is up to developers to explicitly create a correlation id.
The usage would be as follows:
import reconplogger
from flask import Flask
...
app = Flask(__name__)
...
logger = reconplogger.flask_app_logger_setup(app, level='DEBUG')
## NOTE: do not change logger beyond this point!
...
## Use logger in code
myclass = MyClass(..., logger=logger)
...
## User logger in a flask request
@app.route('/')
def hello_world():
logger.info('i like logs')
correlation_id = reconplogger.get_correlation_id()
logger.info('correlation id for this request: '+correlation_id)
return 'Hello, World!'
...
As illustrated in the previous example the :func:`get_correlation_id` function can be used to get the correlation id for the current application context. However, there are cases in which it is desired to set the correlation id, instead of getting a randomly generated one. In this case the :func:`get_correlation_id` function is used, for example as follows:
@app.route('/')
def hello_world():
reconplogger.set_correlation_id('my_correlation_id')
logger.info('i like logs')
return 'Hello, World!'
An important note is that after configuring the logger, the code should not modify the logger configuration. For example, the logging level should not be modified. Adding an additional handler to the logger is not a problem. This could be desired for example to additionally log to a file.
In the helm values.yaml file of the microservice, the default values for the environment variables should be set as:
LOGGER_CFG:
LOGGER_NAME: json_logger
LOGGER_LEVEL: DEBUG
With the json_logger
logger, the format of the logs should look
something like the following:
{"asctime": "2018-09-05 17:38:38,137", "levelname": "INFO", "filename": "test_formatter.py", "lineno": 5, "message": "Hello world"} {"asctime": "2018-09-05 17:38:38,137", "levelname": "DEBUG", "filename": "test_formatter.py", "lineno": 9, "message": "Hello world"} {"asctime": "2018-09-05 17:38:38,137", "levelname": "ERROR", "filename": "test_formatter.py", "lineno": 13, "message": "Hello world"} {"asctime": "2018-09-05 17:38:38,137", "levelname": "CRITICAL", "filename": "test_formatter.py", "lineno": 17, "message": "Hello world"} {"asctime": "2018-09-05 17:38:38,137", "levelname": "ERROR", "filename": "test_formatter.py", "lineno": 25, "message": "division by zero"} {"asctime": "2018-09-05 17:38:38,138", "levelname": "ERROR", "filename": "test_formatter.py", "lineno": 33, "message": "Exception has occured", "exc_info": "Traceback (most recent call last):\n File \"reconplogger/tests/test_formatter.py\", line 31, in test_exception_with_trace\n b = 100 / 0\nZeroDivisionError: division by zero"} {"asctime": "2018-09-05 17:38:38,138", "levelname": "INFO", "filename": "test_formatter.py", "lineno": 37, "message": "Hello world", "context check": "check"} {"asctime": "2020-09-02 17:20:16,428", "levelname": "INFO", "filename": "hello.py", "lineno": 12, "message": "i like logs", "correlation_id": "3958f378-5d48-4e1c-b83b-3c6d9f95faec"} {"asctime": "2020-09-02 17:20:16,428", "levelname": "INFO", "filename": "reconplogger.py", "lineno": 271, "message": "Request is completed", "http_endpoint": "/", "http_method": "GET", "http_response_code": 200, "http_response_size": 56, "http_input_payload_size": null, "http_input_payload_type": null, "http_response_time": "0.0002014636993408203", "correlation_id": "3958f378-5d48-4e1c-b83b-3c6d9f95faec"}
The logger objects returned by the setup functions are normal python
logging.Logger
objects, so all the standard logging functionalities
should be used. Please refer to the logging package documentation for details.
A couple of logging features that should be very commonly used are the
following. To add additional structured information to a log, the extra
argument should be used. A simple example could be:
logger.info('Successfully processed document', extra={'uuid': uuid})
When an exception occurs the exc_info=True
argument should be used, for
example:
try: ... except: logger.critical('Failed to run task', exc_info=True)
In some circumstances it is desired to add to a logger a file handler so that the logging messages are also saved to a file. This normally requires at least three lines of code, thus to simplify things reconplogger provides the :func:`reconplogger.add_file_handler` function to do the same with a single line of code. The use is quite straightforward as:
reconplogger.add_file_handler(logger, '/path/to/log/file.log')
When implementing classes it is common to add logging support to it. For this an
inheritable class :class:`.RLoggerProperty` is included in reconplogger to add an
rlogger
property to easily set and get the reconplogger logger. An
example of use is the following:
from reconplogger import RLoggerProperty
class MyClass(RLoggerProperty):
def __init__(self, logger):
self.rlogger = logger
def my_method(self):
self.rlogger.error('my_method was called')
MyClass(logger=True).my_method()
An important feature of reconplogger is that the logging configuration of apps that use it can be easily changed via the environment variables. First set the environment variables with the desired logging configuration and logger name:
export LOGGER_NAME="example_logger"
export LOGGER_CFG='{
"version": 1,
"formatters": {
"verbose": {
"format": "%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d %(message)s"
}
},
"handlers": {
"console":{
"level":"DEBUG",
"class":"logging.StreamHandler",
"formatter": "verbose"
}
},
"loggers": {
"example_logger": {
"handlers": ["console"],
"level": "ERROR",
}
}
}'
Then, in the python code the logger would be used as follows:
>>> import reconplogger
>>> logger = reconplogger.logger_setup(env_prefix='LOGGER')
>>> logger.error('My error message')
ERROR 2019-10-18 14:45:22,629 <stdin> 16876 139918773925696 My error message
The :func:`reconplogger.load_config` function allows loading of a python logging configuration. The format config can be either json or yaml. The loading of configuration can be from a file (giving its path), from an environment variable (giving the variable name), a raw configuration string, or loading the default configuration that comes with reconplogger. See below examples of loading for each of the cases:
import reconplogger
## Load from config file
reconplogger.load_config('/path/to/config.yaml')
## Load from environment variable
reconplogger.load_config('LOGGER_CFG')
## Load default config
reconplogger.load_config('reconplogger_default_cfg')
In some cases it might be needed to replace the handlers of some already existing logger. For this reconplogger provides the :func:`reconplogger.replace_logger_handlers` function. To use it, simply provide the logger in which to replace the handlers and the logger from where to get the handlers. The procedure would be as follows:
import reconplogger
logger = reconplogger.logger_setup('json_logger')
reconplogger.replace_logger_handlers('some_logger_name', logger)
Contributions to this package are very welcome. When you plan to work with the
source code, note that this project does not include a requirements.txt file.
This is by intention. To make it very clear what are the requirements for
different use cases, all the requirements of the project are stored in the file
setup.cfg. The basic runtime requirements are defined in section
[options]
in the install_requires
entry. All optional
requirements are stored in section [options.extras_require]
. There are
test
, dev
and doc
extras require to be used by
developers (e.g. requirements to run the unit tests) and an all
extras
require for optional runtime requirements, namely Flask support.
The recommended way to work with the source code is the following. First clone the repository, then create a virtual environment, activate it and finally install the development requirements. More precisely the steps would be:
git clone https://github.com/omni-us/reconplogger.git
cd reconplogger
virtualenv -p python3 venv
. venv/bin/activate
The crucial step is installing the requirements which would be done by running:
pip3 install --editable ".[dev]"
Running the unit tests can be done either using using tox or the setup.py
script. The
unit tests are also installed with the package, thus can be used to in a
production system.
tox # Run tests using tox
./setup.py test_coverage # Run tests and generate coverage report
python3 -m reconplogger_tests # Run tests for installed package
To contribute it is required to create and push to a new branch and issue a pull request.
A pull request will only be accepted if:
- All python files pass pylint checks.
- All unit tests run successfully.
- New code has docstrings and gets included in the html documentation.
When developing, after cloning be sure to run the githook-pre-commit to setup the pre-commit hook. This will help you by automatically running pylint before every commit.
Only the maintainer of this repo should bump versions and this should be done
only on the master branch. To bump the version it is required to use the
bumpversion command that should already be installed if pip3 install
--editable .[dev,doc,test,all]
was run as previously instructed.
bumpversion major/minor/patch
Push the tags to the repository as well.
git push; git push --tags
When the version tags are pushed, circleci will automatically build the wheel file, test it and if successful, push the package to pypi.