json-logging

Python logging library to emit JSON log that can be easily indexed and searchable by logging infrastructure such as ELK, EFK, AWS Cloudwatch, GCP Stackdriver

If you're using Cloud Foundry, it might worth to check out the library SAP/cf-python-logging-support which I'm also original author.

Content

1. Features
2. Usage
   2.1 Non-web application log
   2.2 Web application log
   2.3 Get current correlation-id
   2.4 Log extra properties
   2.5 Root logger
   2.6 Custom log formatter
   2.7 Exclude certain URL from request instrumentation
3. Configuration
4. Python References
5. Framework support plugin development
6. FAQ & Troubleshooting
7. References

1. Features

1. Emit JSON logs (format detail)
2. Lightweight, no dependencies, minimal configuration needed (1 LoC to get it working)
3. Seamlessly integrate with Python native logging module. Support both Python 2.7.x and 3.x
4. Auto extract correlation-id for distributed tracing [1]
5. Support HTTP request instrumentation. Built in support for Flask, Sanic, Quart, Connexion. Extensible to support other web frameworks. PR welcome 😃 .
6. Highly customizable: support inject arbitrary extra properties to JSON log message, override logging formatter, etc.
7. Production ready, has been used in production since 2017

2. Usage

Install by running this command:

pip install json-logging

By default log will be emitted in normal format to ease the local development. To enable it on production set enable_json in init_<framework name>(enable_json=True) method call (set json_logging.ENABLE_JSON_LOGGING or ENABLE_JSON_LOGGING environment variable to true is not recommended and will be deprecated in future versions).

To configure, call json_logging.init_< framework_name >(). Once configured library will try to configure all loggers (existing and newly created) to emit log in JSON format.
See following use cases for more detail.

2.1 Non-web application log

This mode don't support correlation-id.

import json_logging, logging, sys

# log is initialized without a web framework name
json_logging.init_non_web(enable_json=True)

logger = logging.getLogger("test-logger")
logger.setLevel(logging.DEBUG)
logger.addHandler(logging.StreamHandler(sys.stdout))

logger.info("test logging statement")

2.2 Web application log

Flask

import datetime, logging, sys, json_logging, flask

app = flask.Flask(__name__)
json_logging.init_flask(enable_json=True)
json_logging.init_request_instrument(app)

# init the logger as usual
logger = logging.getLogger("test-logger")
logger.setLevel(logging.DEBUG)
logger.addHandler(logging.StreamHandler(sys.stdout))

@app.route('/')
def home():
    logger.info("test log statement")
    logger.info("test log statement with extra props", extra={'props': {"extra_property": 'extra_value'}})
    correlation_id = json_logging.get_correlation_id()
    return "Hello world : " + str(datetime.datetime.now())

if __name__ == "__main__":
    app.run(host='0.0.0.0', port=int(5000), use_reloader=False)

Sanic

import logging, sys, json_logging, sanic

app = sanic.Sanic(name="sanic-web-app")
json_logging.init_sanic(enable_json=True)
json_logging.init_request_instrument(app)

# init the logger as usual
logger = logging.getLogger("sanic-integration-test-app")
logger.setLevel(logging.DEBUG)
logger.addHandler(logging.StreamHandler(sys.stdout))

@app.route("/")
async def home(request):
    logger.info("test log statement")
    logger.info("test log statement with extra props", extra={'props': {"extra_property": 'extra_value'}})
    # this will be faster
    correlation_id = json_logging.get_correlation_id(request=request)
    # this will be slower, but will work in context you cant get a reference of request object
    correlation_id_without_request_obj = json_logging.get_correlation_id()

    return sanic.response.text("hello world")

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=8000)

Quart

import asyncio, logging, sys, json_logging, quart

app = quart.Quart(__name__)
json_logging.init_quart(enable_json=True)
json_logging.init_request_instrument(app)

# init the logger as usual
logger = logging.getLogger("test logger")
logger.setLevel(logging.DEBUG)
logger.addHandler(logging.StreamHandler(sys.stdout))

@app.route('/')
async def home():
    logger.info("test log statement")
    logger.info("test log statement with extra props", extra={'props': {"extra_property": 'extra_value'}})
    correlation_id = json_logging.get_correlation_id()
    return "Hello world"

if __name__ == "__main__":
    loop = asyncio.get_event_loop()
    app.run(host='0.0.0.0', port=int(5000), use_reloader=False, loop=loop)

Connexion

import logging, sys, json_logging, connexion

app = connexion.FlaskApp(__name__)
json_logging.init_connexion(enable_json=True)
json_logging.init_request_instrument(app)

app.add_api('api.yaml')

# init the logger as usual
logger = logging.getLogger("test-logger")
logger.setLevel(logging.DEBUG)
logger.addHandler(logging.StreamHandler(sys.stdout))

if __name__ == "__main__":
    app.run()

Custom handler for request instrumentation

you need to explicitly set JSONRequestLogFormatter as default formatter for any extra handler that is added to request_logger

request_logger = json_logging.get_request_logger()
handler = logging.handlers.RotatingFileHandler(filename='log_req.log', maxBytes=5000000, backupCount=10)
handler.setFormatter(json_logging.JSONRequestLogFormatter())
request_logger.addHandler(handler)

2.3 Get current correlation-id

Current request correlation-id can be retrieved and pass to downstream services call as follow:

# this will be faster
correlation_id = json_logging.get_correlation_id(request=request)
# this will be slower, but will work in context where you couldn't get a reference of request object
correlation_id_without_request_obj = json_logging.get_correlation_id()
# use correlation id for downstream service calls here

In request context, if one is not present, a new one might be generated depends on CREATE_CORRELATION_ID_IF_NOT_EXISTS setting value.

2.4 Log extra properties

Extra property can be added to logging statement as follow:

logger.info("test log statement", extra = {'props' : {'extra_property' : 'extra_value'}})

2.5 Root logger

If you want to use root logger as main logger to emit log. Made sure you call config_root_logger() after initialize root logger (by logging.basicConfig() or logging.getLogger()) [2]

logging.basicConfig()
json_logging.init_<framework name >()
json_logging.config_root_logger()

2.6 Custom log formatter

Customer JSON log formatter can be passed to init method. see example for more detail: https://github.com/thangbn/json-logging-python/blob/master/example/custom_log_format.py

2.7 Exclude certain URl from request instrumentation

Certain URL can be excluded from request instrumentation by specifying a list of regex into init_request_instrument method like below:

json_logging.init_request_instrument(app, exclude_url_patterns=[r'/exclude_from_request_instrumentation'])

3. Configuration

logging library can be configured by setting the value in json_logging, all configuration must be placed before json_logging.init method call

4. Python References

TODO: update Python API docs on Github page

5. Framework support plugin development

To add support for a new web framework, you need to extend following classes in framework_base and register support using json_logging.register_framework_support method:

Take a look at json_logging/base_framework.py, json_logging.flask and json_logging.sanic packages for reference implementations.

6. FAQ & Troubleshooting

1. I configured everything, but no logs are printed out?

   • Forgot to add handlers to your logger?
   • Check whether logger is disabled.

2. Same log statement is printed out multiple times.

   • Check whether the same handler is added to both parent and child loggers [2]
   • If you using flask, by default option use_reloader is set to True which will start 2 instances of web application. change it to False to disable this behaviour [3]

7. References

[0] Full logging format references

2 types of logging statement will be emitted by this library:

• Application log: normal logging statement e.g.:

{
	"type": "log",
	"written_at": "2017-12-23T16:55:37.280Z",
	"written_ts": 1514048137280721000,
	"component_id": "1d930c0xd-19-s3213",
	"component_name": "ny-component_name",
	"component_instance_idx": 0,
	"logger": "test logger",
	"thread": "MainThread",
	"level": "INFO",
	"line_no": 22,
	"filename": "/path/to/foo.py"
	"exc_info": "Traceback (most recent call last): \n  File "<stdin>", line 1, in <module>\n ValueError: There is something wrong with your input",
	"correlation_id": "1975a02e-e802-11e7-8971-28b2bd90b19a",
	"extra_property": "extra_value",
	"msg": "This is a message"
}

• Request log: request instrumentation logging statement which recorded request information such as response time, request size, etc.

{
	"type": "request",
	"written_at": "2017-12-23T16:55:37.280Z",
	"written_ts": 1514048137280721000,
	"component_id": "-",
	"component_name": "-",
	"component_instance_idx": 0,
	"correlation_id": "1975a02e-e802-11e7-8971-28b2bd90b19a",
	"remote_user": "user_a",
	"request": "/index.html",
	"referer": "-",
	"x_forwarded_for": "-",
	"protocol": "HTTP/1.1",
	"method": "GET",
	"remote_ip": "127.0.0.1",
	"request_size_b": 1234,
	"remote_host": "127.0.0.1",
	"remote_port": 50160,
	"request_received_at": "2017-12-23T16:55:37.280Z",
	"response_time_ms": 0,
	"response_status": 200,
	"response_size_b": "122",
	"response_content_type": "text/html; charset=utf-8",
	"response_sent_at": "2017-12-23T16:55:37.280Z"
}

See following tables for detail format explanation:

• Common field

• application logs

• request logs:

[1] What is correlation-id/request id

https://stackoverflow.com/questions/25433258/what-is-the-x-request-id-http-header

[2] Python logging propagate

https://docs.python.org/3/library/logging.html#logging.Logger.propagate https://docs.python.org/2/library/logging.html#logging.Logger.propagate

[3] more on flask use_reloader

http://flask.pocoo.org/docs/0.12/errorhandling/#working-with-debuggers