Python logging handlers for sending json-formatted events to loggly. Basically the simplest thing I could make on the quick.
Install via pip:
pip install pyloggly
import logging
from pyloggly import LogglyHandler
logger = logging.getLogger()
handler = LogglyHandler('mytoken', 'logs-01.loggly.com', 'mytag')
logger.addHandler(handler)
logger.info("Hey, I'm logging to loggly!")If you roll that way you can use a logging file config like this:
[handlers]
keys=LogglyHandler
[handler_LogglyHandler]
class=pyloggly.handler.LogglyHandler
args=('mytoken','logs-01.loggly.com','mytag')
[loggers]
keys=root
[logger_root]
handlers=LogglyHandler
level=INFO
[formatters]
keys=
If you want to do something to the response from the loggly api you can pass in a reference to a callback function thusly:
handler = LogglyHandler('token','host','tags', resp_callback=my_callback)
The handler's emit method will catch and re-raise any exceptions of type requests.exceptions.RequestException (and won't catch any others).
You can override this by passing in another callback:
handler = LogglyHandler('token','host','tags', exc_callback=my_callback)
pyloggly also includes a LogglyBulkHandler which utilizes the Loggly bulk api endpoint.
Rather than sending each event is it is emitted, the bulk handler collects events up to batch_size and sends them in batches. A cleanup
function is registered via atexit.register that will send remaining collected events. You can also explicitly call handler.flush().
LogglyHandler uses the standard Loggly HTTPS endpoint and uses a FuturesSession session from requests-futures
to execute async http requests. The handler registers an atexit cleanup function that tries to ensure all pending requests are completed, but YMMV as to reliability.
In some environments, e.g. AWS Lambda, the main program thread may not signal when it exits (note, this is hearsay based on info in another project's README). For those cases it may be necessary to explicitly call the handler's .flush() method.
pyloggly unit tests can be run by executing
python setup.py test
Optionally, you can use tox with the provided tox.ini file
- Jay Luker <jay_luker@harvard.edu> @lbjay, maintainer
Apache 2.0
2018 President and Fellows of Harvard College