Prometheus Metrics in Python — ELI5

Think about a fitness tracker on your wrist.

It constantly measures your heart rate, steps, and sleep. You do not check it every second, but if your heart rate suddenly spikes or your sleep drops, you get an alert. The tracker does not fix the problem — it tells you something is wrong so you can act.

Prometheus works the same way for software. It watches your Python application’s vital signs: how many requests it handles, how fast it responds, how much memory it uses, how many errors happen.

Here is the process:

1. Your Python app keeps a small scoreboard of numbers: “I have handled 5,000 requests. 50 of them failed. The average response time is 200 milliseconds.”
2. Prometheus visits your app every 15 seconds and reads the scoreboard.
3. Prometheus stores all those numbers over time, so you can see trends: “Response time was fine this morning but started climbing at 2 PM.”
4. You set up rules: “If error rate goes above 5%, send me a message.” Prometheus watches for that condition and alerts you.

Real example: an e-commerce site uses Prometheus to monitor its checkout service. One afternoon, the payment provider starts responding slowly. Prometheus notices that the “payment request duration” metric doubled. An alert fires before customers start complaining, and the team switches to a backup provider.

The Python library is called prometheus_client. You add a few lines to your code to create the scoreboard, and Prometheus handles the rest — collection, storage, graphing, and alerting.

One thing to remember: Prometheus is a fitness tracker for your Python app — it constantly measures health numbers and alerts you when something looks wrong, before users notice.