Metrics
The controller exposes Prometheus metrics at GET /metrics (gated by
metrics:read, served at the root path, not under /api/v1). This page lists the
metric names; for the alerting, scrape, and dashboard artifacts see
Observability.
Most series carry labels — the common ones are node_id, channel (and
interface_id) on per-channel audio metrics, severity/status/event_type on
health metrics, action/outcome/permission/actor_type on audit, and
provider/status on uploads. Counters end in _total; everything else is a
gauge.
Controller
| Metric | Meaning |
|---|
rakkr_controller_started_at_seconds | Controller process start timestamp. |
rakkr_database_unavailable | Controller database reachability (1 = unavailable); emitted so /metrics degrades instead of returning 503 during a database outage. |
Nodes
| Metric | Meaning |
|---|
rakkr_node_online | Whether a recorder node is reachable. |
rakkr_node_offline_alerts_active | Unresolved node-offline health events (by node/severity/status). |
Per-channel audio quality, derived from the latest meter frame (labelled by
node_id/interface_id/channel):
| Metric | Meaning |
|---|
rakkr_input_rms_dbfs | Latest RMS input level (dBFS). |
rakkr_input_peak_dbfs | Latest peak input level (dBFS). |
rakkr_input_clipping_ratio | Latest clipping state. |
rakkr_input_speech_score | Speech-likelihood score. |
rakkr_input_noise_score | Non-speech noise score. |
rakkr_input_broadband_noise_score | Broadband-noise likelihood score. |
rakkr_input_hum_score | Hum-likelihood score. |
rakkr_input_static_score | Static-likelihood score. |
rakkr_input_estimated_snr_db | Estimated signal-to-noise ratio (dB). |
rakkr_input_intelligibility_score | First-pass voice intelligibility score. |
rakkr_input_channel_correlation_score | Strongest same-interface channel correlation score (also carries peer_channel + phase labels). |
Live listen monitor
Labelled by node_id/interface_id/channel plus a source label.
| Metric | Meaning |
|---|
rakkr_listen_monitor_chunk_age_seconds | Age of the latest monitor audio chunk. |
rakkr_listen_monitor_chunk_duration_seconds | Duration of the latest chunk. |
Recordings & jobs
| Metric | Meaning |
|---|
rakkr_recording_active | Active recording jobs by node. |
rakkr_recording_cached | Cached recordings by node. |
rakkr_recording_duration_seconds | Recording duration by recording. |
rakkr_recording_bytes_written | Controller-cached recording bytes by recording. |
rakkr_recording_jobs | Recording jobs by node and status. |
rakkr_recording_watchdog_alerts_active | Unresolved watchdog health events (by severity). |
rakkr_recording_watchdog_alerts_total | Watchdog health events raised (by severity). |
rakkr_device_xruns_active | Unresolved audio-xrun health events. |
rakkr_device_xruns_total | Audio-xrun health events (by severity). |
Uploads
| Metric | Meaning |
|---|
rakkr_upload_queue_depth | Upload queue items by provider and status. |
rakkr_upload_queue_oldest_due_seconds | Age of the oldest due queue item. |
rakkr_upload_failures_total | Upload failures. |
Events
| Metric | Meaning |
|---|
rakkr_audit_events_total | Audit events by action, outcome, permission, and actor type. |
rakkr_health_events_active | Unresolved health events by severity and status. |
rakkr_health_events_total | Health events by event type, severity, and status. |
The checked Prometheus alert rules, Mimir remote-write example, and Grafana
dashboard live under docs/observability/.
