{
  "uid": "factory-infra",
  "title": "Factory — Infra",
  "tags": ["factory", "observability", "infra"],
  "timezone": "browser",
  "schemaVersion": 39,
  "version": 1,
  "editable": true,
  "refresh": "30s",
  "time": { "from": "now-6h", "to": "now" },
  "panels": [
    {
      "id": 1,
      "type": "timeseries",
      "title": "Node disk usage (mountpoint /var/lib)",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 0 },
      "fieldConfig": { "defaults": { "unit": "percentunit", "decimals": 1, "thresholds": { "mode": "absolute", "steps": [ { "color": "green", "value": null }, { "color": "red", "value": 0.8 } ] } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" }, "thresholds": { "mode": "absolute" } },
      "targets": [ { "refId": "A", "expr": "node_filesystem_usage{mountpoint=\"/var/lib\"}", "legendFormat": "{{instance}}" } ]
    },
    {
      "id": 2,
      "type": "timeseries",
      "title": "Node CPU usage (1 m)",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 0 },
      "fieldConfig": { "defaults": { "unit": "percentunit", "custom": { "drawStyle": "line" } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" } },
      "targets": [ { "refId": "A", "expr": "1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[1m]))", "legendFormat": "{{instance}}" } ]
    },
    {
      "id": 3,
      "type": "timeseries",
      "title": "Node memory usage",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 8 },
      "fieldConfig": { "defaults": { "unit": "percentunit", "custom": { "drawStyle": "line" } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" } },
      "targets": [ { "refId": "A", "expr": "1 - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes", "legendFormat": "{{instance}}" } ]
    },
    {
      "id": 4,
      "type": "timeseries",
      "title": "DB query p95 per app (10 m) — pool pressure proxy",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 8 },
      "fieldConfig": { "defaults": { "unit": "s", "custom": { "drawStyle": "line" } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" } },
      "targets": [
        { "refId": "A", "expr": "histogram_quantile(0.95, sum by (le) (rate(db_query_duration_seconds_bucket{service_name=\"factory-api\"}[10m])))", "legendFormat": "api" },
        { "refId": "B", "expr": "histogram_quantile(0.95, sum by (le) (rate(db_query_duration_seconds_bucket{service_name=\"factory-worker\"}[10m])))", "legendFormat": "worker" }
      ]
    },
    {
      "id": 5,
      "type": "timeseries",
      "title": "Redis command p95 per app (10 m) — hit-rate pressure proxy",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 0, "y": 16 },
      "fieldConfig": { "defaults": { "unit": "s", "custom": { "drawStyle": "line" } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" } },
      "targets": [
        { "refId": "A", "expr": "histogram_quantile(0.95, sum by (le) (rate(redis_command_duration_seconds_bucket{service_name=\"factory-api\"}[10m])))", "legendFormat": "api" },
        { "refId": "B", "expr": "histogram_quantile(0.95, sum by (le) (rate(redis_command_duration_seconds_bucket{service_name=\"factory-worker\"}[10m])))", "legendFormat": "worker" }
      ]
    },
    {
      "id": 6,
      "type": "timeseries",
      "title": "Storage ops (S3/MinIO) rate + p95 by op",
      "datasource": { "type": "prometheus", "uid": "factory-prometheus" },
      "gridPos": { "h": 8, "w": 12, "x": 12, "y": 16 },
      "fieldConfig": { "defaults": { "unit": "ops", "custom": { "drawStyle": "line" } }, "overrides": [] },
      "options": { "legend": { "displayMode": "list", "placement": "bottom" } },
      "targets": [
        { "refId": "A", "expr": "sum by (op) (rate(storage_ops_total{service_name=~\"factory-(api|worker)\"}[5m]))", "legendFormat": "{{op}} /s" },
        { "refId": "B", "expr": "histogram_quantile(0.95, sum by (le, op) (rate(storage_op_duration_seconds_bucket{service_name=~\"factory-(api|worker)\"}[10m])))", "legendFormat": "{{op}} p95 (s)" }
      ]
    }
  ],
  "annotations": {
    "list": [
      {
        "builtIn": 1,
        "datasource": { "type": "grafana", "uid": "-- Grafana --" },
        "name": "Pod/PG/OS gauges — SPEC GAP",
        "description": "The spec's Infra dashboard names 'CPU/mem per pod, PG connections vs pool, disk, Redis hit rate, OpenSearch cluster health'. Only disk/CPU/mem (node-exporter, expected-present) + app-side proxies (db/redis p95, storage ops) come from the spec's 14-metric set. Per-pod cgroup metrics (kubelet/cadvisor), PG connection pool gauges, redis hit rate and OpenSearch cluster health are EXPORTER-DEPENDENT metrics outside the 14-set — recorded in infrastructure/observability/README.md (spec-conflict, AGENTS rule 29); the panels use the app-side proxies available from the shipped metrics.",
        "enable": true,
        "hide": false,
        "iconColor": "rgba(255, 96, 96, 1)"
      }
    ]
  }
}