As you can imagine, in addition to adding CDK constructs to our stack, we also need to make a few changes to our Flask-based web application so that it starts offering metrics and insights to tools such as Prometheus. Let’s start with the application-level changes first.

Modifying the Flask application code

There are two main changes that we need. First, we must add logging capabilities to the code base, and second, we must allow the code to publish Prometheus metrics that can be consumed by the OpenTelemetry collector that we will deploy later.

Just like other reusable Python modules, Prometheus offers a variety of ready-to-use integrations for commonly used application frameworks. Flask is no different. The two libraries that are relevant for our use case are prometheus-client and prometheus-flask-exporter. It is very simple to expose Prometheus metrics by simply importing these libraries into our code, as we will see shortly. For logging requirements, we will use the standard logging library.

Enabling Prometheus metrics in your application

The Flask application code sits in the app.py file. The complete file is available for your reference athttps://github.com/PacktPublishing/AWS-DevOps-Simplified/blob/ main/chapter-8/chapter-8-flask-app/app.py. Soon after we create an instance of the PrometheusMetrics class, it hooks onto the Flask framework under the hood, thereby directlycapturing all the relevant metrics. Whenever a GET or POST request comes in, Prometheus gets to know about it. This avoids any extra effort from the developer’s side unless they want to define and export additional custom metrics based on the application’s business use case:

…from prometheus_flask_exporterimport PrometheusMetrics……app = Flask(__name__)metrics = PrometheusMetrics(app)…

To include these modules in our build workflow, we should also add them to the requirements. txt file so that they are picked up when a Docker image is created. As a best practice, we should lockthem down with exact version identifiers:

……Werkzeug==2.2.3zipp==3.14.0

prometheus-client==0.16.0 prometheus-flask-exporter==0.22.3

Deploying the application at this point will automatically enable the /metrics endpoint, which exports Prometheus metrics extracted from the Flask application.

This completes all the changes we require in the application code. You can build your own Docker image at this point, or use the DockerHub version, available at akskap/flask-todo-list-app-with-metrics, which includes all the changes already.

Extending the CDK stack with new constructs

Now that we have made the code changes for our Python application, let’s define the additional infrastructure elements we need in the CDK stack.

Note

All the code definitions are available for your reference at https://github.com/PacktPublishing/AWS-DevOps-Simplified/blob/main/chapter-8/ chapter-8-cdk/lib/chapter-8-cdk-stack.ts.

As usual, all the CDK construct definitions can be found inside the lib/ folder of the project directory, chapter-8/chapter-8-cdk/, hosted in your Cloud9 IDE.