We use Sentry to monitor errors within our Django applications. It’s an excellent tool: you should use it.

Regrettable however, we sometimes need to SSH into a server within our platform and use Django’s shell to explore or adjust data. However, by default, exceptions from these sessions are captured up Sentry and appear in the dashboard.

This isn’t helpful as such errors are not application problems per se and just add noise. We want Sentry to ignore these exceptions.

To prevent these errors being captured, the solution is to use a custom Sentry client class with an overridden should_capture method that ignores errors triggered from a shell session.

Here’s an example:

import sys

from raven.contrib.django import DjangoClient


class CustomSentryClient(DjangoClient):

    def should_capture(self, exc_info):
        # Check if this exception was triggered from a shell session. We don't care about these
        # as they are normally human typos and of no further interest.
        if len(sys.argv) >= 2:
            if sys.argv[1] in ("shell", "shell_plus", "dbshell"):
                return False

        return super().should_capture(exc_info)

Plumb this in by adding a SENTRY_CLIENT setting specifying the module path to this class:

# settings.py

SENTRY_CLIENT = 'path.to.module.CustomSentryClient'

Problem solved.

David Winterbottom

Head of Engineering