DataDog Logger

Note: This library does not send logs to Datadog's Log Management product. See Datadog's documentation for how to configure log collection. See https://docs.datadoghq.com/logs/log_collection/

A Python logging.Handler for sending log messages to DataDog as Events in the Events Explorer.

Installation

pip install datadog-logger

Usage

The simplest way to enable logging to DataDog is to use the log_error_events helper, which will cause all logging.ERROR and higher messages to be sent to DataDog:

import datadog
from datadog_logger import log_error_events
import logging

# Authenticate with DataDog
datadog.initialize(api_key="api-key", app_key="app-key")

# Note, a normal STDOUT handler will not be configured if this is not
# called first
logging.basicConfig()

log_error_events(tags=["tag1:value", "tag2:value"], mentions=["@devs", "@slack"])

logging.error("Oh no!")

This will enable the handler on the root logger; tags and mentions may both be None, in which case none will be included. The handler will have a log level of logging.ERROR, meaning only ERROR and CRITICAL (or any higher custom levels) will be sent to DataDog.

A specific logger may also be specified by name:

log_error_events("some.logger")

Details

The Handler may also be created and added to a logger manually:

import datadog
from datadog_logger import DatadogLogHandler
import logging

datadog.initialize(api_key="api-key", app_key="app-key")

datadog_handler = DatadogLogHandler(
    tags=["some:tag"], mentions=["@some-mention"], level=logging.WARNING)

# Enable STDOUT logging
logging.basicConfig()

# The root logger
logger = logging.getLogger()
logger.addHandler(datadog_handler)

# This will be logged, because the level is WARNING
logger.warning("Watch out!")

# This will not
logger.info("Lets not log everything to DataDog")

When the event is created, the msg of the log record will be passed as the title of the event. Normally, this would be the string passed into the logging call, with any % substitutions applied.

The text of the event will be the formatted record. If the log record includes exc_info (i.e. because you called logging.exception, or passed exc_info to the log function), then the full stack trace will be included in the event text.

Any @mentions passed via the mentions constructor argument will be appended to the end of the text, so this library can be used to alert, via DataDog, on particular log messages.