Skip to content

Health Endpoints

xp-tracker exposes health and readiness endpoints for Kubernetes probes.

GET /healthz

Liveness probe. Always returns 200 OK with body ok.

Use this to detect if the process is alive. The HTTP server binds on startup and responds immediately -- no dependencies on Kubernetes API or store state.

GET /readyz

Readiness probe. Returns 503 Service Unavailable until the first poll cycle completes, then 200 OK with body ok.

This prevents Kubernetes from sending traffic (including Prometheus scrapes) to the exporter before it has populated the in-memory store. Without this, Prometheus would scrape empty metrics on startup.

Deployment probes

The base Deployment manifests configure both probes:

livenessProbe:
  httpGet:
    path: /healthz
    port: metrics
  initialDelaySeconds: 5
  periodSeconds: 10
readinessProbe:
  httpGet:
    path: /readyz
    port: metrics
  initialDelaySeconds: 5
  periodSeconds: 10

Tip

If your first poll cycle takes longer than the default readiness probe timeout (e.g., large number of GVRs or slow API server), increase initialDelaySeconds or POLL_INTERVAL_SECONDS accordingly.