titan-iac/services/monitoring/grafana-dashboard-public.yaml

# services/monitoring/grafana-dashboard-public.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: grafana-dashboard-public
  labels:
    grafana_dashboard: "1"
data:
  atlas-public-overview.json: |
    {
      "annotations": {
        "list": [
          {
            "builtIn": 1,
            "datasource": {
              "type": "datasource",
              "uid": "grafana"
            },
            "enable": true,
            "hide": true,
            "iconColor": "rgba(0, 211, 255, 1)",
            "name": "Annotations & Alerts",
            "type": "dashboard"
          }
        ]
      },
      "editable": false,
      "folderUid": "atlas-public",
      "graphTooltip": 0,
      "links": [],
      "panels": [
        {
          "id": 1,
          "type": "stat",
          "title": "Running pods",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 6,
            "w": 6,
            "x": 0,
            "y": 0
          },
          "targets": [
            {
              "expr": "sum(kube_pod_status_phase{phase=\"Running\"})",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "absolute",
                "steps": [
                  {
                    "color": "rgba(115, 115, 115, 1)",
                    "value": null
                  },
                  {
                    "color": "green",
                    "value": 1
                  }
                ]
              },
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 2,
          "type": "stat",
          "title": "Ready node percentage",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 6,
            "w": 6,
            "x": 6,
            "y": 0
          },
          "targets": [
            {
              "expr": "sum(kube_node_status_condition{condition=\"Ready\",status=\"true\"}) / sum(kube_node_info) * 100",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "percentage",
                "steps": [
                  {
                    "color": "red",
                    "value": null
                  },
                  {
                    "color": "orange",
                    "value": 90
                  },
                  {
                    "color": "green",
                    "value": 98
                  }
                ]
              },
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 3,
          "type": "stat",
          "title": "Cluster CPU saturation",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 6,
            "w": 6,
            "x": 12,
            "y": 0
          },
          "targets": [
            {
              "expr": "avg((1 - rate(node_cpu_seconds_total{mode=\"idle\"}[5m])) * 100)",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "percentage",
                "steps": [
                  {
                    "color": "green",
                    "value": null
                  },
                  {
                    "color": "yellow",
                    "value": 65
                  },
                  {
                    "color": "red",
                    "value": 85
                  }
                ]
              },
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 4,
          "type": "stat",
          "title": "Cluster memory usage",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 6,
            "w": 6,
            "x": 18,
            "y": 0
          },
          "targets": [
            {
              "expr": "100 - (sum(node_memory_MemAvailable_bytes) / sum(node_memory_MemTotal_bytes) * 100)",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "percentage",
                "steps": [
                  {
                    "color": "green",
                    "value": null
                  },
                  {
                    "color": "yellow",
                    "value": 70
                  },
                  {
                    "color": "red",
                    "value": 85
                  }
                ]
              },
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 5,
          "type": "piechart",
          "title": "Namespace CPU share",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 9,
            "w": 12,
            "x": 0,
            "y": 6
          },
          "targets": [
            {
              "expr": "topk(8, sum(rate(container_cpu_usage_seconds_total{namespace!=\"\",pod!=\"\",container!=\"\"}[5m])) by (namespace))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "cores"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "list",
              "placement": "right"
            },
            "pieType": "pie",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 6,
          "type": "piechart",
          "title": "Namespace memory share",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 9,
            "w": 12,
            "x": 12,
            "y": 6
          },
          "targets": [
            {
              "expr": "topk(8, sum(container_memory_working_set_bytes{namespace!=\"\",pod!=\"\",container!=\"\"}) by (namespace))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "bytes"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "list",
              "placement": "right"
            },
            "pieType": "donut",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 7,
          "type": "timeseries",
          "title": "Node CPU usage (per node)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 0,
            "y": 15
          },
          "targets": [
            {
              "expr": "avg(rate(node_cpu_seconds_total{mode!=\"idle\"}[5m])) by (instance) * 100",
              "refId": "A",
              "legendFormat": "{{instance}}"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 8,
          "type": "timeseries",
          "title": "Node memory usage (per node)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 12,
            "y": 15
          },
          "targets": [
            {
              "expr": "avg((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100) by (instance)",
              "refId": "A",
              "legendFormat": "{{instance}}"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 9,
          "type": "table",
          "title": "Key service availability",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 0,
            "y": 23
          },
          "targets": [
            {
              "expr": "max by (service) (up{service=~\"traefik|gitea|vault|victoria-metrics-single|grafana|alertmanager\"})",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "mappings": [
                {
                  "id": 0,
                  "type": 1,
                  "value": "0",
                  "text": "Down"
                },
                {
                  "id": 1,
                  "type": 1,
                  "value": "1",
                  "text": "Up"
                }
              ],
              "thresholds": {
                "mode": "absolute",
                "steps": [
                  {
                    "color": "red",
                    "value": null
                  },
                  {
                    "color": "green",
                    "value": 1
                  }
                ]
              }
            },
            "overrides": []
          },
          "options": {
            "showHeader": true
          }
        },
        {
          "id": 10,
          "type": "table",
          "title": "Failed pods (24h trend)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 12,
            "y": 23
          },
          "targets": [
            {
              "expr": "topk(10, sum(increase(kube_pod_status_phase{phase=\"Failed\"}[24h])) by (namespace))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "showHeader": true
          }
        },
        {
          "id": 11,
          "type": "timeseries",
          "title": "Cluster network throughput",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 0,
            "y": 31
          },
          "targets": [
            {
              "expr": "sum(rate(container_network_receive_bytes_total{namespace!=\"\"}[5m]))",
              "refId": "A",
              "legendFormat": "Receive"
            },
            {
              "expr": "sum(rate(container_network_transmit_bytes_total{namespace!=\"\"}[5m]))",
              "refId": "B",
              "legendFormat": "Transmit"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "Bps"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 12,
          "type": "timeseries",
          "title": "Storage usage across nodes",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 12,
            "y": 31
          },
          "targets": [
            {
              "expr": "100 - (sum(node_filesystem_avail_bytes{mountpoint=\"/\",fstype!~\"tmpfs|overlay\"}) / sum(node_filesystem_size_bytes{mountpoint=\"/\",fstype!~\"tmpfs|overlay\"}) * 100)",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "list",
              "placement": "bottom"
            }
          }
        },
        {
          "id": 13,
          "type": "text",
          "title": "About this dashboard",
          "gridPos": {
            "h": 6,
            "w": 24,
            "x": 0,
            "y": 39
          },
          "options": {
            "content": "### Atlas at a glance\n- Raspberry Pi + Jetson hybrid cluster with Flux-managed GitOps\n- Metrics powered by VictoriaMetrics, visualized by Grafana\n- Login for SRE mode with pod-level drilldowns, alert routes, and storage health",
            "mode": "markdown"
          }
        }
      ],
      "refresh": "30s",
      "schemaVersion": 39,
      "style": "dark",
      "tags": [
        "atlas",
        "public"
      ],
      "templating": {
        "list": []
      },
      "time": {
        "from": "now-12h",
        "to": "now"
      },
      "title": "Atlas Public Overview",
      "uid": "atlas-public",
      "version": 3
    }