Skip to content
Raddex

Metrics

Start raddex with --metrics-addr to expose Prometheus metrics over HTTP:

Terminal window
raddex run -c Raddexfile --metrics-addr 127.0.0.1:9100

The metrics listener serves Prometheus text format at /metrics:

Terminal window
curl http://127.0.0.1:9100/metrics

Metrics

Section titled “Metrics”

All metrics are registered on the default Prometheus registry and recorded on every completed request:

Metric Type Description
raddex_requests_total Counter Total HTTP requests served by raddex (the QPS source)
raddex_request_duration_seconds Histogram HTTP request duration in seconds 
# HELP raddex_requests_total Total HTTP requests served by raddex
# TYPE raddex_requests_total counter
raddex_requests_total 123456


# HELP raddex_request_duration_seconds HTTP request duration in seconds
# TYPE raddex_request_duration_seconds histogram
raddex_request_duration_seconds_bucket{le="0.005"} 90000
raddex_request_duration_seconds_bucket{le="0.01"} 120000
raddex_request_duration_seconds_sum 620.5
raddex_request_duration_seconds_count 123456

The histogram uses the default Prometheus buckets. Compute latency percentiles (p50 / p99) from the buckets with the usual histogram_quantile() PromQL:

histogram_quantile(0.99, sum by (le) (rate(raddex_request_duration_seconds_bucket[5m])))

What is not measured yet

Section titled “What is not measured yet”

The metrics set is deliberately minimal. There are no per-site or per-upstream breakdowns, and no backend connection-pool gauges yet. OpenTelemetry tracing is on the roadmap.