Prometheus metrics, health checks for orchestrators, and structured logs.
The metrics listener (global.metrics_addr) serves two paths: /metrics (Prometheus text format) and /healthz (returns 200 ok; wire it into liveness and readiness probes).
| Metric | Type | Meaning |
|---|---|---|
http_requests_total | counter | every request received |
http_request_duration_seconds | histogram | time from request receipt to upstream response headers |
upstream_connection_errors | counter | failed or timed-out upstream requests |
scrape_configs: - job_name: torana static_configs: - targets: ["localhost:9090"]
Structured JSON on stdout by default (log_format = "json"); set any other value for human-readable text. Levels follow log_level. Startup logs every bound listener; warnings fire for accepted-but-unimplemented config fields.
Next: deployment recipes.