Profile a Python App with Datadog's Performance Profiler Agent

1. 1
   Install Dependencies

   Install `dd-trace-py`, Datadog's tracing and profiling library for Python, along with Flask to create a sample web application to profile.
2. 2
   Create a Sample Application

   Create a simple Python Flask application with a CPU-intensive endpoint. This gives the Performance Profiler Agent something meaningful to analyze. Save the code provided in the 'Starter' section as `app.py`.
3. 3
   Run with the Profiler Agent

   Instead of `python app.py`, use `ddtrace-run`. This command automatically instruments your application. Set environment variables to enable profiling and tag the data with your service name, environment, and version.
4. 4
   Generate Load

   The profiler collects data when your application is active. Use a simple loop with `curl` in a separate terminal to continuously hit your `/heavy` endpoint and generate performance data.
5. 5
   Analyze CPU Flame Graphs

   Navigate to 'APM > Profiling' in your Datadog account. Select your service (`my-profiler-app`). Here you can analyze the CPU flame graph to find which functions consume the most resources. Wider bars represent more time spent.

# 1. Save this code as app.py
from flask import Flask
import time

app = Flask(__name__)

# This function simulates a CPU-intensive task
def cpu_intensive_task():
    result = 0
    for i in range(10_000_000):
        result += i
    return result

@app.route('/')
def index():
    return "Welcome to the profiling demo app!"

@app.route('/heavy')
def heavy_request():
    start_time = time.time()
    cpu_intensive_task()
    end_time = time.time()
    duration = end_time - start_time
    return f"Heavy task completed in {duration:.2f} seconds."

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=5001)

# --- How to Run ---
# 2. Install dependencies:
# pip install Flask dd-trace-py

# 3. Run the application with the Datadog profiler enabled.
#    (Requires a Datadog Agent running on the host or container)
#
# export DD_SERVICE="my-profiler-app"
# export DD_ENV="development"
# export DD_VERSION="1.0.0"
# export DD_PROFILING_ENABLED="true"
#
# ddtrace-run python app.py

# 4. In a separate terminal, send requests to generate load:
# while true; do curl http://localhost:5001/heavy; echo; sleep 1; done

# 5. Go to your Datadog account: APM > Profiling.
#    Select the "my-profiler-app" service to view flame graphs.