Добавлены метрики Prometheus и дашборды Grafana (этап 2.5)

Все три Go-сервиса (ingest, device-control, rule-engine) отдают /metrics в формате Prometheus: счётчики MQTT/RabbitMQ/ClickHouse операций, гистограммы длительности батч-флашей и диспатча правил, переходы устройств online/offline. Добавлены сервисы prometheus и grafana в docker-compose с провижининг конфигом (datasource + два готовых дашборда: "Ingest & Telemetry" и "Automation & Devices").
2026-08-11 21:58:59 +05:00
parent 3ade5c2512
commit 81040eec62
31 changed files with 712 additions and 53 deletions
@@ -0,0 +1,83 @@
+{
+  "uid": "automation-devices",
+  "title": "Automation & Devices",
+  "tags": ["home-automation"],
+  "timezone": "browser",
+  "schemaVersion": 39,
+  "version": 1,
+  "refresh": "10s",
+  "time": { "from": "now-30m", "to": "now" },
+  "panels": [
+    {
+      "id": 1,
+      "title": "Rules triggered (rate, by outcome)",
+      "type": "timeseries",
+      "datasource": { "type": "prometheus", "uid": "Prometheus" },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
+      "targets": [
+        {
+          "expr": "sum by (outcome) (rate(rule_engine_rules_triggered_total[1m]))",
+          "legendFormat": "{{outcome}}",
+          "refId": "A"
+        }
+      ]
+    },
+    {
+      "id": 2,
+      "title": "device-control-service dispatch duration (p95, by action)",
+      "type": "timeseries",
+      "datasource": { "type": "prometheus", "uid": "Prometheus" },
+      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 0 },
+      "fieldConfig": { "defaults": { "unit": "s" }, "overrides": [] },
+      "targets": [
+        {
+          "expr": "histogram_quantile(0.95, sum(rate(rule_engine_dispatch_duration_seconds_bucket[5m])) by (le, action_type))",
+          "legendFormat": "{{action_type}}",
+          "refId": "A"
+        }
+      ]
+    },
+    {
+      "id": 3,
+      "title": "Device commands (rate, by action + outcome)",
+      "type": "timeseries",
+      "datasource": { "type": "prometheus", "uid": "Prometheus" },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
+      "targets": [
+        {
+          "expr": "sum by (action, outcome) (rate(device_control_commands_total[1m]))",
+          "legendFormat": "{{action}} / {{outcome}}",
+          "refId": "A"
+        }
+      ]
+    },
+    {
+      "id": 4,
+      "title": "Device online/offline transitions (rate)",
+      "type": "timeseries",
+      "datasource": { "type": "prometheus", "uid": "Prometheus" },
+      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 8 },
+      "targets": [
+        {
+          "expr": "sum by (direction) (rate(device_control_health_transitions_total[5m]))",
+          "legendFormat": "{{direction}}",
+          "refId": "A"
+        }
+      ]
+    },
+    {
+      "id": 5,
+      "title": "Automation rule cache refreshes (rate, by outcome)",
+      "type": "timeseries",
+      "datasource": { "type": "prometheus", "uid": "Prometheus" },
+      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 16 },
+      "targets": [
+        {
+          "expr": "sum by (outcome) (rate(rule_engine_cache_refresh_total[1m]))",
+          "legendFormat": "{{outcome}}",
+          "refId": "A"
+        }
+      ]
+    }
+  ]
+}