Troubleshooting ‘failed to find attribute’ errors on Python Linux App Services

5 minute read | By Anthony Salemo

In this post we’ll cover troubleshooting “failed to find attribute ‘app’ in ‘app’” based messages for Python on App Service Linux.

Overview

failed to find attribute 'app' in 'app' is an error thrown back by Gunicorn letting us know it cannot find the wsgi or asgi callable that it expects to find in the code that you’re deploying.

Although, this can be any combination of names (like ‘test’ in ‘somefile’), as it’s based on the reasons we cover later.

Python Blessed Images on App Service Linux use Oryx to run applications, by default. Therefor, the startup command, unless otherwise configured, is slightly opinionated on what it expects to run by default.

Oryx uses the following logic to find this callable’s file, which expects to be in the root of your project being deployed:

1. If user has specified a start script, run it.
2. Else, find a WSGI module and run with gunicorn.
    i. Look for and run a directory containing a wsgi.py file (for Django).
    ii. Look for the following files in the root of the repo and an app class within them (for Flask and other WSGI frameworks).
        - application.py
        - app.py
        - index.py
        - server.py

What is a wsgi or asgi callable?

Taking Flask and an example, it is simply the following:

app = Flask(__name__)

Per Flasks documentation:

The flask object implements a WSGI application and acts as the central object. It is passed the name of the module or package of the application. Once it is created it will act as a central registry for the view functions, the URL rules, template configuration and much more.

Or, with FastAPI:

app = FastAPI()

This is more or less the same for most wSGI-based (Flask, Django, Hug, Dash) and aSGI-based applications (FastAPI, Quart).

Why does this happen?

As mentioned above, this is because Gunicorn cannot find the wSGI or aSGI callable. This means either:

  • The callable is not named app
  • The file in which the callable lives is not in the root of your project
  • If relying on Oryx for the startup command, then your callable file name is not one of the following: application.py, app.py, index.py or server.py (in addition to points 1 and 2)
  • Note: For Django, Oryx expects the callable to be named application, which would be in mysite/wsgi.py in a typical Django structure. If is not, it’ll manifest as Failed to find attribute 'application' in 'mysite.wsgi'.
    • If your wsgi.py file is more than one (1) folder deep, Oryx will fail to find the location and show No framework detected

Oryx runs a Gunicorn command on startup that is generally like this:

gunicorn --bind 0.0.0.0 --timeout 600 --access-logfile '-' --error-logfile '-' app:app

The contributor to this error is the last entry in this particular command, which equates to:

gunicorn --bind 0.0.0.0 --timeout 600 --access-logfile '-' --error-logfile '-' <the_file_that_has_your_callable>:<your_callable_variable_name>

app:app would mean:

  • app.py - in the current directory this is executing from (Left side of the colon (:))
  • app = SomeCallable() - the variable name of the callable within app.py (Right side of the colon (:))

Take another example, for instance:

gunicorn --bind 0.0.0.0 --timeout 600 --access-logfile '-' --error-logfile '-' application:call

application:call would mean:

  • application.py - in the current directory this is executing from (Left side of the colon (:))
  • call = SomeCallable() - the variable name of the callable within app.py (Right side of the colon (:))

If your project was set up like this, it would surface as Failed to find attribute 'app' in 'application'., or the following, if the file name doesn’t match any of the ones Oryx expects:

No framework detected; using default app from /opt/defaultsite
Generating `gunicorn` command for 'application:app'

Resolution

If using Oryx-provided startup

If you’re relying on Oryx (which means not having a custom startup command), then ensure you’re following the criteria covered above and within the Oryx Python run detector documentation.

If using a custom startup command

If you’re using a custom startup command, or, needing to use one - such as if have a wSGI/aSGI file not in the root or in a nested folder, or a non-standard Oryx named file, ensure the startup command is updated appropriately.

Example - non-root directory

Take the below file structure as an example.

| -- .gitignore
| -- README.md
| -- requirements.txt
| -- parentdir
    | -- app.py

This would obviously fail to run with Oryx’s current logic. Therefor we’d have to update our Gunicorn startup command to the following:

gunicorn --bind 0.0.0.0 --timeout 600 --access-logfile '-' --error-logfile '-' --chdir parentdir app:app

This is changing the directory to run the callable under parentdir and then executing against app.py, assuming the callable itself is named app.

Example - targetting different file and module names

Take the below file structure as an example.

| -- .gitignore
| -- README.md
| -- requirements.txt
| -- appservice.py

Where appservice.py contains the following callable based on Dash:

lib = dash.Dash(__name__, server=server, external_stylesheets=external_stylesheets)

This would need to have a custom startup command set to the following:

gunicorn --bind 0.0.0.0 --timeout 600 --access-logfile '-' --error-logfile '-' appservice:lib

Example - Django

For Django specific examples and explanation on changing your command for either the callable file name or callable itself, see here.

You may also enjoy