Observability

Effective observability is crucial for managing dynamic preview environments. Diverge is designed to be highly observable, providing operators with clear visibility into the system’s state.

Core Principles

1. Native Metrics: Diverge exposes metrics in the Prometheus exposition format.
2. Controller Visibility: Standard controller-runtime metrics are available for the operator.
3. Health Checking: Granular endpoints for liveness, readiness, and controller status.

Component Monitoring

• Controller: Monitors API interactions, reconcile durations, queue depth, async route lifecycle, and per-environment cost estimation (diverge_environment_estimated_hourly_cost_usd labeled by environment and namespace).
• Server: Monitors RPC requests, latency, authentication attempts, and active streams.
• Proxy & Activator: Tracks routing decisions, cold-start wake-up events, and last-traffic timestamps (divergedev.com/last-traffic-at) used by the controller to enforce idle hibernation (divergedev.com/hibernate-after).

These metrics enable teams to build comprehensive Grafana dashboards, attribute preview infrastructure spend per pull request, and set up reliable alerting rules.