Observability Stack

A runnable monitoring lab: a Python service exposes Prometheus metrics, Prometheus scrapes it, and Grafana displays a provisioned dashboard.

Start the stack

docker compose up --build

Open the application at http://localhost:8000, Prometheus at http://localhost:9090, and Grafana at http://localhost:3000. The dashboard and data source are provisioned automatically.

Portfolio highlights

• metrics endpoint and scrape configuration;
• Docker Compose service dependencies and read-only configuration mounts;
• Grafana dashboard as code;
• Kubernetes Deployment, Service and Prometheus Operator ServiceMonitor;
• CI checks tests, JSON configuration and image build.

Replace OWNER in k8s/app.yaml before pushing the image. The ServiceMonitor requires the Prometheus Operator or kube-prometheus-stack.

Architecture

App (/metrics) -> Prometheus (scrape) -> Grafana (provisioned dashboard)
                                    -> Prometheus Operator ServiceMonitor (k8s)

Key files

• app/main.py — exposes app_uptime_seconds and serves the app.
• prometheus/prometheus.yml — scrape config; grafana/provisioning/* — dashboard and data source as code.
• compose.yaml — service dependencies and read-only config mounts.

What I learned

"Dashboards as code" means the Grafana provisioning YAML is version controlled and reviewed like everything else, so a broken dashboard is a merge conflict, not a manual click. Sharing the metrics port across the Compose stack and giving Prometheus a service dependency ordering keeps the scrape loop stable during startup, and the same manifest works unchanged for the k8s ServiceMonitor.

Ports

Service	Port	Use
App	8000	http://localhost:8000
Prometheus	9090	metrics browser
Grafana	3000	dashboards
(Prod) ServiceMonitor	cluster-IP	cluster scraping