Skip to content

KPIs & Dashboards

The dashboard is the operational heart of LogSys: a live view of system health driven by KPI aggregates served from the KPIs API and pushed over WebSocket.

KPI Data Path

flowchart LR
  PG[(PostgreSQL<br/>materialized views)] --> KV[KPI service]
  KV --> CACHE[(Redis cache<br/>kpis_global:*)]
  KV --> API[GET /api/kpis]
  CACHE --> API
  API --> REST[REST poll]
  API --> WS[WS /ws/kpis]
  REST --> UI[React dashboard]
  WS --> UI
  • compute_kpis reads materialized views + live counts
  • Results cached in Redis (kpis_global:*), invalidated on new events/clear-all
  • WebSocket sends kpi:snapshot on connect and when the cache changes

KPI Cards

Card Source Refresh
Total events mv_dashboard_aggregates snapshot
Total occurrences sum(occurrence_count) snapshot
Error events (is_error_or_worse) matview filter snapshot
Events last minute events_last_min snapshot
Criticals 24h criticals_24h snapshot
Priority counts (P1–P4) events + template_priority snapshot
DLQ count dlq_events snapshot

Charts

Chart Data
Severity breakdown GET /api/analytics/severity-breakdown
Top sources GET /api/analytics/top-sources
Time series GET /api/logs/meta/timeseries?window=…
Calendar GET /api/logs/meta/calendar
DLQ vs stored GET /api/analytics/dlq-vs-stored

Real-time Behavior

  1. On mount, useKpiStream connects to /ws/kpis
  2. Server sends initial kpi:snapshot — dashboard renders instantly
  3. When new events are written, the pipeline publishes events:new on Redis; KPI cache invalidates; next snapshot reflects it
  4. Connection loss → auto-reconnect with backoff; interim REST polls keep data fresh

View Selectors

DashboardViewSelector lets operators switch between:

  • Live view — real-time, WebSocket driven
  • Range view — historical aggregates (REST)

Empty-State Handling

If no events exist (e.g. simulator stopped), charts show empty states and the time series endpoint returns an empty buckets array with has_more: false — no 500s, no spinner loops.