Duplicate error log for "HTTP 500 Internal Server Error"

[browniebroke]

• Django DataDog Logger version: 0.6.3
• Python version: 3.12
• Operating System: Docker python:3.12-slim

Description

In case of 5XX errors, the RequestLoggingMiddleware logs an error, using logger.error:

django-datadog-logger/django_datadog_logger/middleware/request_log.py

Lines 38 to 44 in d8a88da

	elif 500 <= response.status_code < 600:
	log_entry_dict["error.kind"] = response.status_code
	log_entry_dict["error.message"] = response.reason_phrase
	logger.error(
	f"HTTP {response.status_code} {response.reason_phrase}",
	extra=log_entry_dict,
	)

However, I noticed that it duplicates error logging. In case of 500 error, it seems like Django already logs the exception, and with more context. This causes all errors to be reported twice: once with the generic Django handler and once with this middleware.

We're using Sentry and I noticed that the error logged from the middleware is less actionable: Sentry groups all occurrences together, and there is no stack trace attached.

I was going to suggest to add the option to override the action for error logging in subclasses, so that I could customise it more easily, but maybe we shouldn't log anything in the case of a 500, given that Django logging is good enough.

Then I thought a about more and wondered if anything >=500 make sense?

• 501: I'm never used this status, but I guess once could choose to return that
• 502, 503, 504: would typically be raised by a reverse proxy in front of the server, not by Django itself
• 505 and above: I'm not too sure to be honest...

What I Did

• Got a 500 error
• Saw 2 errors in Datadog (and Sentry) instead of one