mirror of
https://github.com/ced4568/cedops-ai-infrastructure-noc.git
synced 2026-08-12 21:34:02 +00:00
1555 lines
34 KiB
JSON
1555 lines
34 KiB
JSON
{
|
|
"annotations": {
|
|
"list": []
|
|
},
|
|
"editable": true,
|
|
"fiscalYearStartMonth": 0,
|
|
"graphTooltip": 1,
|
|
"id": null,
|
|
"links": [],
|
|
"panels": [
|
|
{
|
|
"id": 1,
|
|
"type": "row",
|
|
"title": "Executive Reliability Overview",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 0
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 2,
|
|
"type": "stat",
|
|
"title": "Synthos VPS",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 4,
|
|
"x": 0,
|
|
"y": 1
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "up{instance=\"synthos-node-1\"}",
|
|
"legendFormat": "Synthos VPS",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 1
|
|
}
|
|
]
|
|
},
|
|
"mappings": [
|
|
{
|
|
"type": "value",
|
|
"options": {
|
|
"0": {
|
|
"text": "DOWN",
|
|
"color": "red"
|
|
},
|
|
"1": {
|
|
"text": "UP",
|
|
"color": "green"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 3,
|
|
"type": "stat",
|
|
"title": "NOC Health Score",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 5,
|
|
"x": 4,
|
|
"y": 1
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "100 * avg(probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"})",
|
|
"legendFormat": "NOC Health Score",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 80
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 95
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 4,
|
|
"type": "stat",
|
|
"title": "24h Public SLA",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 5,
|
|
"x": 9,
|
|
"y": 1
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "100 * avg(avg_over_time(probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"}[24h]))",
|
|
"legendFormat": "24h Public SLA",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 95
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 99
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 5,
|
|
"type": "stat",
|
|
"title": "Active Alerts",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 5,
|
|
"x": 14,
|
|
"y": 1
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum(ALERTS{alertstate=\"firing\"}) OR vector(0)",
|
|
"legendFormat": "Active Alerts",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 1
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 3
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 6,
|
|
"type": "stat",
|
|
"title": "Min SSL Days",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 5,
|
|
"x": 19,
|
|
"y": 1
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "min((probe_ssl_earliest_cert_expiry{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"} - time()) / 86400)",
|
|
"legendFormat": "Min SSL Days",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "d",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 14
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 30
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 7,
|
|
"type": "row",
|
|
"title": "VPS Capacity & Host Health",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 5
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 8,
|
|
"type": "timeseries",
|
|
"title": "VPS CPU Usage %",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 8,
|
|
"x": 0,
|
|
"y": 6
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "100 - (avg by(instance)(rate(node_cpu_seconds_total{instance=\"synthos-node-1\",mode=\"idle\"}[5m])) * 100)",
|
|
"legendFormat": "CPU",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 70
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 90
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 9,
|
|
"type": "timeseries",
|
|
"title": "VPS Memory Usage %",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 8,
|
|
"x": 8,
|
|
"y": 6
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "(1 - (node_memory_MemAvailable_bytes{instance=\"synthos-node-1\"} / node_memory_MemTotal_bytes{instance=\"synthos-node-1\"})) * 100",
|
|
"legendFormat": "Memory",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 75
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 90
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 10,
|
|
"type": "timeseries",
|
|
"title": "VPS Disk Usage %",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 8,
|
|
"x": 16,
|
|
"y": 6
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "100 - ((node_filesystem_avail_bytes{instance=\"synthos-node-1\",mountpoint=\"/\",fstype!=\"rootfs\"} * 100) / node_filesystem_size_bytes{instance=\"synthos-node-1\",mountpoint=\"/\",fstype!=\"rootfs\"})",
|
|
"legendFormat": "Disk /",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 70
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 85
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 11,
|
|
"type": "timeseries",
|
|
"title": "VPS Network Traffic",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 13
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum by(instance)(rate(node_network_receive_bytes_total{instance=\"synthos-node-1\",device!~\"lo|veth.*|docker.*|br-.*\"}[5m]))",
|
|
"legendFormat": "RX",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"expr": "sum by(instance)(rate(node_network_transmit_bytes_total{instance=\"synthos-node-1\",device!~\"lo|veth.*|docker.*|br-.*\"}[5m]))",
|
|
"legendFormat": "TX",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "Bps",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 12,
|
|
"type": "timeseries",
|
|
"title": "VPS Load Average",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 13
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "node_load1{instance=\"synthos-node-1\"}",
|
|
"legendFormat": "1m",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"expr": "node_load5{instance=\"synthos-node-1\"}",
|
|
"legendFormat": "5m",
|
|
"refId": "B"
|
|
},
|
|
{
|
|
"expr": "node_load15{instance=\"synthos-node-1\"}",
|
|
"legendFormat": "15m",
|
|
"refId": "C"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 13,
|
|
"type": "row",
|
|
"title": "Public Services: Availability, Latency, Status",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 20
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 14,
|
|
"type": "stat",
|
|
"title": "Public Services Up",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 0,
|
|
"y": 21
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum(probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"})",
|
|
"legendFormat": "Public Services Up",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 3
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 5
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 15,
|
|
"type": "stat",
|
|
"title": "Public Services Total",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 6,
|
|
"y": 21
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "count(probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"})",
|
|
"legendFormat": "Public Services Total",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 16,
|
|
"type": "stat",
|
|
"title": "Avg Response Time",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 12,
|
|
"y": 21
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "avg(probe_duration_seconds{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"})",
|
|
"legendFormat": "Avg Response Time",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 1
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 3
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 17,
|
|
"type": "stat",
|
|
"title": "HTTP 2xx Services",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 18,
|
|
"y": 21
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "count(probe_http_status_code{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"} >= 200 <= 299)",
|
|
"legendFormat": "HTTP 2xx Services",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 3
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 5
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 18,
|
|
"type": "timeseries",
|
|
"title": "Public Endpoint Response Time",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 25
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "probe_duration_seconds{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"}",
|
|
"legendFormat": "{{service}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 1
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 3
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 19,
|
|
"type": "timeseries",
|
|
"title": "HTTP Status Codes",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 25
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "probe_http_status_code{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"}",
|
|
"legendFormat": "{{service}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 20,
|
|
"type": "row",
|
|
"title": "SSL / SLA / Reliability",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 33
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 21,
|
|
"type": "timeseries",
|
|
"title": "SSL Certificate Days Remaining",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 34
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "(probe_ssl_earliest_cert_expiry{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"} - time()) / 86400",
|
|
"legendFormat": "{{service}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "d",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 14
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 30
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 22,
|
|
"type": "timeseries",
|
|
"title": "24h Service SLA %",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 8,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 34
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "avg_over_time(probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"}[24h]) * 100",
|
|
"legendFormat": "{{service}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 95
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 99
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 23,
|
|
"type": "timeseries",
|
|
"title": "Probe Failure Timeline",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 42
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "1 - probe_success{job=\"blackbox-http-public\",site=~\"synthos-systems|ceds-home-lab\"}",
|
|
"legendFormat": "{{service}} failure",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 24,
|
|
"type": "row",
|
|
"title": "Prometheus / Scrape Health",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 49
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 25,
|
|
"type": "stat",
|
|
"title": "Targets Down",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 0,
|
|
"y": 50
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "count(up == 0)",
|
|
"legendFormat": "Targets Down",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 1
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 3
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 26,
|
|
"type": "stat",
|
|
"title": "Node Exporters Up",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 6,
|
|
"y": 50
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum(up{job=~\".*node-exporter.*\"})",
|
|
"legendFormat": "Node Exporters Up",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 3
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 6
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 27,
|
|
"type": "stat",
|
|
"title": "Blackbox Public Up",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 12,
|
|
"y": 50
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "sum(probe_success{job=\"blackbox-http-public\"})",
|
|
"legendFormat": "Blackbox Public Up",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "red",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 3
|
|
},
|
|
{
|
|
"color": "green",
|
|
"value": 5
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 28,
|
|
"type": "stat",
|
|
"title": "TSDB Head Series",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 4,
|
|
"w": 6,
|
|
"x": 18,
|
|
"y": 50
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "prometheus_tsdb_head_series",
|
|
"legendFormat": "TSDB Head Series",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
},
|
|
{
|
|
"color": "yellow",
|
|
"value": 100000
|
|
},
|
|
{
|
|
"color": "red",
|
|
"value": 250000
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 29,
|
|
"type": "timeseries",
|
|
"title": "Scrape Duration by Job",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 54
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "scrape_duration_seconds",
|
|
"legendFormat": "{{job}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 30,
|
|
"type": "timeseries",
|
|
"title": "Samples Scraped by Job",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "${DS_PROMETHEUS}"
|
|
},
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 54
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "scrape_samples_scraped",
|
|
"legendFormat": "{{job}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "table",
|
|
"placement": "bottom"
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 31,
|
|
"type": "row",
|
|
"title": "Ops Notes / Cost Awareness",
|
|
"gridPos": {
|
|
"h": 1,
|
|
"w": 24,
|
|
"x": 0,
|
|
"y": 61
|
|
},
|
|
"collapsed": false,
|
|
"panels": []
|
|
},
|
|
{
|
|
"id": 32,
|
|
"type": "text",
|
|
"title": "CedOps Operational Notes",
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 0,
|
|
"y": 62
|
|
},
|
|
"options": {
|
|
"mode": "markdown",
|
|
"content": "### Operating Model\n\n- **V1** = executive NOC overview\n- **V2** = platform engineering drill-down\n- VPS metrics are scraped over **Tailscale**, not the public internet\n- Public services are monitored through **Blackbox Exporter**\n- Designed for portfolio, interview, and operations use\n\n### Next Improvements\n\n- Telegram/Discord alert routing\n- Container-level metrics via cAdvisor\n- Synthetic checks for Paperclip/n8n workflows\n- Incident timeline and MTTR tracking"
|
|
}
|
|
},
|
|
{
|
|
"id": 33,
|
|
"type": "text",
|
|
"title": "Cost + Capacity Awareness",
|
|
"gridPos": {
|
|
"h": 7,
|
|
"w": 12,
|
|
"x": 12,
|
|
"y": 62
|
|
},
|
|
"options": {
|
|
"mode": "markdown",
|
|
"content": "### Cost-Aware Platform Notes\n\nTrack whether the VPS is right-sized:\n\n- CPU sustained over 70% = review workload sizing\n- Memory sustained over 75% = review containers/models\n- Disk over 70% = plan cleanup or expansion\n- Public latency over 1s = investigate routing/app performance\n\nThis section shows business-aware platform engineering: reliability, cost, and capacity together."
|
|
}
|
|
}
|
|
],
|
|
"refresh": "30s",
|
|
"schemaVersion": 39,
|
|
"style": "dark",
|
|
"tags": [
|
|
"cedops",
|
|
"synthos",
|
|
"prometheus",
|
|
"blackbox",
|
|
"platform-engineering",
|
|
"v2"
|
|
],
|
|
"templating": {
|
|
"list": [
|
|
{
|
|
"current": {},
|
|
"hide": 0,
|
|
"includeAll": false,
|
|
"label": "Prometheus",
|
|
"multi": false,
|
|
"name": "DS_PROMETHEUS",
|
|
"options": [],
|
|
"query": "prometheus",
|
|
"refresh": 1,
|
|
"regex": "",
|
|
"type": "datasource"
|
|
}
|
|
]
|
|
},
|
|
"time": {
|
|
"from": "now-24h",
|
|
"to": "now"
|
|
},
|
|
"timepicker": {},
|
|
"timezone": "browser",
|
|
"title": "CedOps AI Infrastructure NOC V2",
|
|
"uid": "cedops-ai-infra-noc-v2",
|
|
"version": 1,
|
|
"weekStart": ""
|
|
} |