v0.3 - Real Data Collection

Originally, I had this application returning mock data to confirm the workflow. Now, it's actually querying the Envoy Proxy sidecars to collect information.

Real Metrics Collection (pkg/istio/discovery.go:209-382)

The application uses genuine Kubernetes API calls to collect real metrics:

Real Kubernetes API Interaction: Lines 217-244 show it executes kubectl exec commands inside actual pods to access Istio's Envoy proxy metrics endpoint (http://localhost:15020/stats/prometheus)

Actual Envoy Metrics Parsing: Lines 258-382 parse real Prometheus-format metrics from Envoy proxies, including:

• istio_requests_total for request counts
• istio_request_duration_milliseconds for latency percentiles
• envoy_http_downstream_cx_active for connection counts
• Circuit breaker, retry, and timeout metrics

Live Pod Discovery: Lines 113-156 discover services by scanning actual Kubernetes pods with Istio sidecars using real pod labels and annotations

Kubernetes Client Integration: The application connects to your actual cluster using kubeconfig (pkg/k8s/client.go) and performs live API calls

Real Data Flow

From the scan command (cmd/scan.go:111), the flow is:

1. Connect to a real Kubernetes cluster
2. Discover actual services with Istio sidecars
3. Execute real curl commands inside Envoy sidecar containers
4. Parse real Prometheus metrics from Envoy
5. Apply ML clustering and anomaly detection to actual data

v0.2

Removed the learn functionality as it was no longer needed.

v0.1

Initial release. Contains the following features:

• smanalyzer scan: One-time anomaly scan
• smanalyzer learn: Train baseline behavior model
• smanalyzer monitor: Continuous real-time monitoring
• smanalyzer status: System health and configuration overview