Django: Adding a Custom Attribute to All Log Lines

Adrienne Domingus
3 min readMay 1, 2018

Django provides logging out of the box, and gives us the tools to customize what gets logged how.

This is not an in-depth primer on logging with Django— I’ve linked to the relevant documentation for the different pieces discussed here, but this is more an end-to-end implementation than a deep-dive — we’re going to look at how to add a custom attribute to all of our log lines. For example, if we have a multi-tenant application, we may want to know which organization performed each action for which there is a log line. So let’s implement this, by adding an organization_id to each log line.

Adding an attribute to a local thread

In order to add an attribute to our log lines, the attribute will first need to be accessible on the local thread — by the time we get to the point of altering our log record, we don’t have access to the request, but we do have access to the thread. Adding an attribute can be accomplished with middleware, in which we can add a custom attribute in process_request. We then remove it in process_response — by the time we get there the log lines have already been processed and it is no longer necessary. Here’s what a simple middleware class to add the organization_id to the thread local might look like:

(The below is based on Django v. 1.10 — The syntax of this has changed with Django 2.0, so you may need to adjust using the documentation if you’re running a newer version.)

import threadinglocal = threading.local()
class OrganizationIdMiddleware(object):
def process_request(self, request):
organization_id = # logic to get attribute off request
setattr(local, 'organization_id', organization_id)

def process_response(self, request, response):
setattr(local, 'organization_id', None)
return response

We’ll then add our new middleware class to so that it gets run for each request/response cycle:



We’ll now create a custom logging filter, to get the attribute off of the thread and set it on the log record, so that it can be used later. At the most basic, our new filter might look something like this:

import logging

Recommended from Medium


See more recommendations