# Python instrumentation:

from prometheus_client import Counter, Histogram, start_http_server

requests_total = Counter("myapp_requests_total","Total requests",["status"])

latency = Histogram("myapp_request_duration_seconds","Request latency")

# Expose metrics port in Service:

ports: [{name: metrics, containerPort: 8081}]

# ServiceMonitor CRD (Prometheus Operator picks this up):

spec:

selector: {matchLabels: {app: myapp}}

endpoints: [{port: metrics, interval: 30s}]

# Verify: kubectl port-forward svc/prometheus 9090 → check /targets

ServiceMonitor is the GitOps-friendly pattern: each team deploys their own ServiceMonitor alongside their Deployment. Prometheus auto-discovers it without central config changes. Metric naming conventions: myapp_http_requests_total (counter), myapp_request_duration_seconds (histogram), myapp_active_connections (gauge). Unit in name, snake_case, _total suffix for counters.