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": 5,
            "w": 4,
            "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
            },
            "textMode": "value"
          }
        },
        {
          "id": 2,
          "type": "stat",
          "title": "Ready nodes",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 5,
            "w": 4,
            "x": 4,
            "y": 0
          },
          "targets": [
            {
              "expr": "sum(kube_node_status_condition{condition=\\\"Ready\\\",status=\\\"true\\\"})",
              "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
            },
            "textMode": "value"
          }
        },
        {
          "id": 3,
          "type": "stat",
          "title": "Cluster nodes",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 5,
            "w": 4,
            "x": 8,
            "y": 0
          },
          "targets": [
            {
              "expr": "count(kube_node_info)",
              "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
            },
            "textMode": "value"
          }
        },
        {
          "id": 4,
          "type": "stat",
          "title": "Hottest node CPU",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 5,
            "w": 4,
            "x": 12,
            "y": 0
          },
          "targets": [
            {
              "expr": "topk(1, avg by (node) ((((1 - label_replace(rate(node_cpu_seconds_total{mode=\\\"idle\\\"}[5m]), \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\")) * 100) * on (internal_ip) group_left(node) kube_node_info)))",
              "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
            },
            "textMode": "value_and_name"
          }
        },
        {
          "id": 5,
          "type": "stat",
          "title": "Hottest node memory",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 5,
            "w": 4,
            "x": 16,
            "y": 0
          },
          "targets": [
            {
              "expr": "topk(1, avg by (node) (((label_replace((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes, \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\") * 100) * on (internal_ip) group_left(node) kube_node_info)))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "percentage",
                "steps": [
                  {
                    "color": "green",
                    "value": null
                  },
                  {
                    "color": "yellow",
                    "value": 75
                  },
                  {
                    "color": "red",
                    "value": 90
                  }
                ]
              },
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            },
            "textMode": "value_and_name"
          }
        },
        {
          "id": 6,
          "type": "stat",
          "title": "Failed pods (24h)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 5,
            "w": 4,
            "x": 20,
            "y": 0
          },
          "targets": [
            {
              "expr": "sum(increase(kube_pod_status_phase{phase=\\\"Failed\\\"}[24h]))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "color": {
                "mode": "palette-classic"
              },
              "mappings": [],
              "thresholds": {
                "mode": "absolute",
                "steps": [
                  {
                    "color": "green",
                    "value": null
                  },
                  {
                    "color": "yellow",
                    "value": 1
                  },
                  {
                    "color": "red",
                    "value": 3
                  }
                ]
              },
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            },
            "textMode": "value"
          }
        },
        {
          "id": 7,
          "type": "piechart",
          "title": "Namespace CPU share",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 9,
            "w": 12,
            "x": 0,
            "y": 5
          },
          "targets": [
            {
              "expr": "topk(10, 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": 8,
          "type": "piechart",
          "title": "Namespace memory share",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 9,
            "w": 12,
            "x": 12,
            "y": 5
          },
          "targets": [
            {
              "expr": "topk(10, 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": 9,
          "type": "timeseries",
          "title": "Node CPU usage (per node)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 0,
            "y": 14
          },
          "targets": [
            {
              "expr": "avg by (node) ((((1 - label_replace(rate(node_cpu_seconds_total{mode=\\\"idle\\\"}[5m]), \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\")) * 100) * on (internal_ip) group_left(node) kube_node_info))",
              "refId": "A",
              "legendFormat": "{{node}}"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 10,
          "type": "timeseries",
          "title": "Node memory usage (per node)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 12,
            "y": 14
          },
          "targets": [
            {
              "expr": "avg by (node) (((label_replace((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes, \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\") * 100) * on (internal_ip) group_left(node) kube_node_info))",
              "refId": "A",
              "legendFormat": "{{node}}"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 11,
          "type": "table",
          "title": "Key service availability",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 12,
            "x": 0,
            "y": 22
          },
          "targets": [
            {
              "expr": "label_replace((sum by (deployment,namespace) (kube_deployment_status_replicas_available{deployment=~\\\"traefik|gitea|grafana\\\",namespace=~\\\"traefik|gitea|monitoring\\\"}) / sum by (deployment,namespace) (kube_deployment_spec_replicas{deployment=~\\\"traefik|gitea|grafana\\\",namespace=~\\\"traefik|gitea|monitoring\\\"})), \\\"service\\\", \\\"$1\\\", \\\"deployment\\\", \\\"(.*)\\\") or label_replace((sum by (statefulset,namespace) (kube_statefulset_status_replicas_ready{statefulset=~\\\"vault|alertmanager|victoria-metrics-single-server\\\",namespace=~\\\"vault|monitoring\\\"}) / sum by (statefulset,namespace) (kube_statefulset_status_replicas{statefulset=~\\\"vault|alertmanager|victoria-metrics-single-server\\\",namespace=~\\\"vault|monitoring\\\"})), \\\"service\\\", \\\"$1\\\", \\\"statefulset\\\", \\\"(.*)\\\")",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "custom": {
                "align": "auto"
              },
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "showHeader": true
          }
        },
        {
          "id": 12,
          "type": "table",
          "title": "Failed pods by namespace (24h)",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 12,
            "x": 12,
            "y": 22
          },
          "targets": [
            {
              "expr": "topk(10, sum by (namespace) (increase(kube_pod_status_phase{phase=\\\"Failed\\\"}[24h])))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "showHeader": true
          }
        },
        {
          "id": 13,
          "type": "timeseries",
          "title": "Root filesystem usage per node",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 0,
            "y": 29
          },
          "targets": [
            {
              "expr": "avg by (node) ((label_replace(1 - (node_filesystem_avail_bytes{mountpoint=\\\"/\\\",fstype!~\\\"tmpfs|overlay\\\"} / node_filesystem_size_bytes{mountpoint=\\\"/\\\",fstype!~\\\"tmpfs|overlay\\\"}), \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\") * 100) * on (internal_ip) group_left(node) kube_node_info)",
              "refId": "A",
              "legendFormat": "{{node}}"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "table",
              "placement": "bottom"
            },
            "tooltip": {
              "mode": "multi"
            }
          }
        },
        {
          "id": 14,
          "type": "bargauge",
          "title": "Nodes closest to full root disks",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 8,
            "w": 12,
            "x": 12,
            "y": 29
          },
          "targets": [
            {
              "expr": "topk(8, avg by (node) ((label_replace(1 - (node_filesystem_avail_bytes{mountpoint=\\\"/\\\",fstype!~\\\"tmpfs|overlay\\\"} / node_filesystem_size_bytes{mountpoint=\\\"/\\\",fstype!~\\\"tmpfs|overlay\\\"}), \"internal_ip\", \"$1\", \"instance\", \"([^:]+):.*\") * 100) * on (internal_ip) group_left(node) kube_node_info))",
              "refId": "A"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "percent",
              "min": 0,
              "max": 100
            },
            "overrides": []
          },
          "options": {
            "displayMode": "gradient",
            "orientation": "horizontal",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 15,
          "type": "stat",
          "title": "Astreae usage",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 6,
            "x": 0,
            "y": 37
          },
          "targets": [
            {
              "expr": "(sum(longhorn_disk_usage_bytes{disk=~\\\"astreae-.*\\\"}) / sum(longhorn_disk_capacity_bytes{disk=~\\\"astreae-.*\\\"})) * 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
            },
            "textMode": "value"
          }
        },
        {
          "id": 16,
          "type": "stat",
          "title": "Asteria usage",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 6,
            "x": 6,
            "y": 37
          },
          "targets": [
            {
              "expr": "(sum(longhorn_disk_usage_bytes{disk=~\\\"asteria-.*\\\"}) / sum(longhorn_disk_capacity_bytes{disk=~\\\"asteria-.*\\\"})) * 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
            },
            "textMode": "value"
          }
        },
        {
          "id": 17,
          "type": "stat",
          "title": "Astreae schedulable",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 6,
            "x": 12,
            "y": 37
          },
          "targets": [
            {
              "expr": "(sum(longhorn_disk_capacity_bytes{disk=~\\\"astreae-.*\\\"}) - sum(longhorn_disk_usage_bytes{disk=~\\\"astreae-.*\\\"}) - sum(longhorn_disk_reservation_bytes{disk=~\\\"astreae-.*\\\"}))",
              "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": "bytesSI"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            },
            "textMode": "value"
          }
        },
        {
          "id": 18,
          "type": "stat",
          "title": "Asteria schedulable",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 6,
            "x": 18,
            "y": 37
          },
          "targets": [
            {
              "expr": "(sum(longhorn_disk_capacity_bytes{disk=~\\\"asteria-.*\\\"}) - sum(longhorn_disk_usage_bytes{disk=~\\\"asteria-.*\\\"}) - sum(longhorn_disk_reservation_bytes{disk=~\\\"asteria-.*\\\"}))",
              "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": "bytesSI"
            },
            "overrides": []
          },
          "options": {
            "colorMode": "value",
            "graphMode": "area",
            "justifyMode": "center",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            },
            "textMode": "value"
          }
        },
        {
          "id": 19,
          "type": "piechart",
          "title": "Longhorn node readiness",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 12,
            "x": 0,
            "y": 44
          },
          "targets": [
            {
              "expr": "sum(longhorn_node_status{condition=\\\"ready\\\"})",
              "refId": "A",
              "legendFormat": "Ready"
            },
            {
              "expr": "(longhorn_node_count_total - sum(longhorn_node_status{condition=\\\"ready\\\"}))",
              "refId": "B",
              "legendFormat": "Offline"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "list",
              "placement": "right"
            },
            "pieType": "donut",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 20,
          "type": "piechart",
          "title": "Longhorn disk schedulability",
          "datasource": {
            "type": "prometheus",
            "uid": "atlas-vm"
          },
          "gridPos": {
            "h": 7,
            "w": 12,
            "x": 12,
            "y": 44
          },
          "targets": [
            {
              "expr": "sum(sum by (node,disk) (longhorn_disk_status{condition=\\\"schedulable\\\"}))",
              "refId": "A",
              "legendFormat": "Schedulable"
            },
            {
              "expr": "(count(sum by (node,disk) (longhorn_disk_status{condition=\\\"ready\\\"})) - sum(sum by (node,disk) (longhorn_disk_status{condition=\\\"schedulable\\\"})))",
              "refId": "B",
              "legendFormat": "Blocked"
            }
          ],
          "fieldConfig": {
            "defaults": {
              "unit": "none"
            },
            "overrides": []
          },
          "options": {
            "legend": {
              "displayMode": "list",
              "placement": "right"
            },
            "pieType": "donut",
            "reduceOptions": {
              "calcs": [
                "lastNotNull"
              ],
              "fields": "",
              "values": false
            }
          }
        },
        {
          "id": 21,
          "type": "text",
          "title": "About this dashboard",
          "gridPos": {
            "h": 5,
            "w": 24,
            "x": 0,
            "y": 51
          },
          "options": {
            "mode": "markdown",
            "content": "### Atlas at a glance\\n- Flux-managed Pi + Jetson cluster with 20+ active nodes\\n- Longhorn tiers: Astreae (3x replicas) & Asteria (2x replicas) tracked separately\\n- Login for SRE mode with alert routing, Longhorn drilldowns, and workload burn rates"
          }
        }
      ],
      "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": 5
    }