Metrics

The operator, the command sidecar in every Play pod, the idle screen, and media-api each serve Prometheus metrics on :9200, under milestone 65 ’s shared contract: every process on the cluster network uses the same port, since no two of these pods ever share one. The base in deploy/ needs no Prometheus: every listener answers with nothing scraping it. An owner who runs the prometheus-operator adds deploy/monitoring, the Component that carries a PodMonitor for each of these pods, beside the base.

Component Metric Type Why
media-operator media_reconcile_duration_seconds{kind} histogram how long one reconcile of one kind took
media-operator media_reconcile_errors_total{kind} counter reconciles that returned an error
media-operator media_watch_restarts_total{kind} counter a watch’s stream ended and the operator opened it again
media-operator media_players{zone, state} gauge idle, playing, paused
media-operator media_playback_starts_total counter plays over time
media-operator media_playback_failures_total{reason} counter the film did not start
media-operator media_display_restarts_total{player} counter the film plays with no on-screen display
media-operator, idle-screen media_bus_connected gauge the MQTT bus is up
command sidecar media_decode_info{codec, hardware} gauge, info a software decode is a 1300 millicore line
command sidecar media_dropped_frames_total counter stutter
command sidecar media_av_delay_seconds gauge sync drift
media-api media_api_requests_total{route, method, status} counter requests by route and outcome
media-api media_api_request_seconds{route} histogram header time
media-api media_api_streams_active{aspect} gauge captures open now
media-api media_api_upstream_requests_total{upstream, status} counter what display-api and audio-api answered
media-api media_api_compose_offset_seconds histogram the measured header offset
media-api media_api_compose_offset_last_seconds gauge the last correction, signed: negative when the audio headers arrived first
media-api media_api_certificate_expiry_seconds gauge the serving leaf runs out
resources:
  - https://github.com/liken-sh/liken//media-operator/deploy?ref=<tag>
components:
  - https://github.com/liken-sh/liken//media-operator/deploy/monitoring?ref=<tag>