Predictive Reliability for a Microservice

Overview

This project tends to be an example of solutions used in site reliabiity combined with data science.

I came up with this idea a bit by accident and finished the whole project in one evening. I simply wanted to learn the basics of Kubernetes, as I have never worked with it before. I watched a couple of tutorials and then asked myself - what if I tried to simulate traffic on a website and do a small project related to site reliability? That is how it came to life.

The workspace is quite simple: Traffic generator > Microservice (Fast API) > Prometheus Metrics > Grafana > (query API) > ML Pipeline (anomaly, forecasts) > (alerts) > Slack/WebHook

What I Learned

• Setting up Grafana, Prometheus, Docker-compose
• Generating random trafic spikes on a website
• Predicting and analysing the data with Python AI libraries

Technical Challenges

• None

Future Improvements

• Adding Slack/Webhook
• Use of K8s