Django Instrumentation

This shows how to use opentelemetry-instrumentation-django to automatically instrument a Django app.

For more user convenience, a Django app is already provided in this directory.

Preparation

This example will be executed in a separate virtual environment:

$ mkdir django_auto_instrumentation
$ virtualenv django_auto_instrumentation
$ source django_auto_instrumentation/bin/activate

Installation

$ pip install opentelemetry-sdk
$ pip install opentelemetry-instrumentation-django
$ pip install requests

Execution

Execution of the Django app

This example uses Django features intended for development environment. The runserver option should not be used for production environments.

Set these environment variables first:

  1. export DJANGO_SETTINGS_MODULE=instrumentation_example.settings

The way to achieve OpenTelemetry instrumentation for your Django app is to use an opentelemetry.instrumentation.django.DjangoInstrumentor to instrument the app.

Clone the opentelemetry-python repository and go to opentelemetry-python/docs/examples/django.

Once there, open the manage.py file. The call to DjangoInstrumentor().instrument() in main is all that is needed to make the app be instrumented.

Run the Django app with python manage.py runserver --noreload. The --noreload flag is needed to avoid Django from running main twice.

Execution of the client

Open up a new console and activate the previous virtual environment there too:

source django_auto_instrumentation/bin/activate

Go to opentelemetry-python/docs/examples/django, once there run the client with:

python client.py hello

Go to the previous console, where the Django app is running. You should see output similar to this one:

{
    "name": "home_page_view",
    "context": {
        "trace_id": "0xed88755c56d95d05a506f5f70e7849b9",
        "span_id": "0x0a94c7a60e0650d5",
        "trace_state": "{}"
    },
    "kind": "SpanKind.SERVER",
    "parent_id": "0x3096ef92e621c22d",
    "start_time": "2020-04-26T01:49:57.205833Z",
    "end_time": "2020-04-26T01:49:57.206214Z",
    "status": {
        "status_code": "OK"
    },
    "attributes": {
        "http.method": "GET",
        "http.server_name": "localhost",
        "http.scheme": "http",
        "host.port": 8000,
        "http.host": "localhost:8000",
        "http.url": "http://localhost:8000/?param=hello",
        "net.peer.ip": "127.0.0.1",
        "http.flavor": "1.1",
        "http.status_text": "OK",
        "http.status_code": 200
    },
    "events": [],
    "links": []
}

The last output shows spans automatically generated by the OpenTelemetry Django Instrumentation package.

Disabling Django Instrumentation

Django’s instrumentation can be disabled by setting the following environment variable:

export OTEL_PYTHON_DJANGO_INSTRUMENT=False

Auto Instrumentation

This same example can be run using auto instrumentation. Comment out the call to DjangoInstrumento().instrument() in main, then Run the django app with opentelemetry-instrument python manage.py runserver --noreload. Repeat the steps with the client, the result should be the same.

Usage with Auto Instrumentation and uWSGI

uWSGI and Django can be used together with auto instrumentation. To do so, first install uWSGI in the previous virtual environment:

pip install uwsgi

Once that is done, run the server with uwsgi from the directory that contains instrumentation_example:

opentelemetry-instrument uwsgi --http :8000 --module instrumentation_example.wsgi

This should start one uWSGI worker in your console. Open up a browser and point it to localhost:8000. This request should display a span exported in the server console.