pez-infra/ansible/services/grafana/dashboards/infrastructure.json
Rasmus Wejlgaard 737d6e0bc1 initial commit
2026-03-28 12:39:41 +00:00

1034 lines
25 KiB
JSON

{
"annotations": {
"list": [
{
"builtIn": 1,
"datasource": {
"type": "datasource",
"uid": "grafana"
},
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
"name": "Annotations & Alerts",
"target": {
"limit": 100,
"matchAny": false,
"tags": [],
"type": "dashboard"
},
"type": "dashboard"
}
]
},
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 1,
"links": [
{
"icon": "external link",
"tags": [],
"targetBlank": true,
"title": "GitHub",
"type": "link",
"url": "https://github.com/rfmoz/grafana-dashboards"
},
{
"icon": "external link",
"tags": [],
"targetBlank": true,
"title": "Grafana",
"type": "link",
"url": "https://grafana.com/grafana/dashboards/1860"
}
],
"panels": [
{
"collapsed": false,
"gridPos": {
"h": 1,
"w": 24,
"x": 0,
"y": 0
},
"id": 261,
"panels": [],
"repeat": "nodename",
"title": "",
"type": "row"
},
{
"fieldConfig": {
"defaults": {},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 0,
"y": 1
},
"id": 324,
"options": {
"code": {
"language": "plaintext",
"showLineNumbers": false,
"showMiniMap": false
},
"content": "<table style=\"width:100%; height:100%;border:0px solid black;\">\n <td style=\"text-align: center;vertical-align: middle;border:0px solid black; \"><h1>${nodename}</h1></td>\n</table>",
"mode": "html"
},
"pluginVersion": "12.1.0",
"title": "",
"type": "text"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "Resource pressure via PSI",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"links": [],
"mappings": [],
"max": 1,
"min": 0,
"thresholds": {
"mode": "percentage",
"steps": [
{
"color": "green",
"value": 0
},
{
"color": "dark-yellow",
"value": 70
},
{
"color": "dark-red",
"value": 90
}
]
},
"unit": "percentunit"
},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 3,
"y": 1
},
"id": 323,
"options": {
"displayMode": "basic",
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": false
},
"maxVizHeight": 300,
"minVizHeight": 10,
"minVizWidth": 0,
"namePlacement": "auto",
"orientation": "horizontal",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showUnfilled": true,
"sizing": "auto",
"text": {},
"valueMode": "color"
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "irate(node_pressure_cpu_waiting_seconds_total{server=\"${nodename}\",job=\"$job\"}[$__rate_interval])",
"format": "time_series",
"instant": true,
"legendFormat": "CPU",
"range": false,
"refId": "A",
"step": 240
},
{
"editorMode": "code",
"exemplar": false,
"expr": "irate(node_pressure_memory_waiting_seconds_total{server=\"${nodename}\",job=\"$job\"}[$__rate_interval])",
"format": "time_series",
"instant": true,
"legendFormat": "Mem",
"range": false,
"refId": "B",
"step": 240
},
{
"editorMode": "code",
"exemplar": false,
"expr": "irate(node_pressure_io_waiting_seconds_total{server=\"${nodename}\",job=\"$job\"}[$__rate_interval])",
"format": "time_series",
"instant": true,
"legendFormat": "I/O",
"range": false,
"refId": "C",
"step": 240
},
{
"editorMode": "code",
"exemplar": false,
"expr": "irate(node_pressure_irq_stalled_seconds_total{server=\"${nodename}\",job=\"$job\"}[$__rate_interval])",
"format": "time_series",
"instant": true,
"legendFormat": "Irq",
"range": false,
"refId": "D",
"step": 240
}
],
"title": "Pressure",
"type": "bargauge"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "Overall CPU busy percentage (averaged across all cores)",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"max": 100,
"min": 0,
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "rgba(50, 172, 45, 0.97)",
"value": 0
},
{
"color": "rgba(237, 129, 40, 0.89)",
"value": 85
},
{
"color": "rgba(245, 54, 54, 0.9)",
"value": 95
}
]
},
"unit": "percent"
},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 6,
"y": 1
},
"id": 20,
"options": {
"minVizHeight": 75,
"minVizWidth": 75,
"orientation": "auto",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showThresholdLabels": false,
"showThresholdMarkers": true,
"sizing": "auto"
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "100 * (1 - avg(rate(node_cpu_seconds_total{mode=\"idle\", server=\"${nodename}\"}[$__rate_interval])))",
"instant": true,
"legendFormat": "",
"range": false,
"refId": "A",
"step": 240
}
],
"title": "CPU Busy",
"type": "gauge"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "System load over all CPU cores together",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"max": 100,
"min": 0,
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "rgba(50, 172, 45, 0.97)",
"value": 0
},
{
"color": "rgba(237, 129, 40, 0.89)",
"value": 85
},
{
"color": "rgba(245, 54, 54, 0.9)",
"value": 95
}
]
},
"unit": "percent"
},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 9,
"y": 1
},
"id": 155,
"options": {
"minVizHeight": 75,
"minVizWidth": 75,
"orientation": "auto",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showThresholdLabels": false,
"showThresholdMarkers": true,
"sizing": "auto"
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "scalar(node_load1{server=\"${nodename}\",job=\"$job\"}) * 100 / count(count(node_cpu_seconds_total{server=\"${nodename}\",job=\"$job\"}) by (cpu))",
"format": "time_series",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "Sys Load",
"type": "gauge"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "Real RAM usage excluding cache and reclaimable memory",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"mappings": [],
"max": 100,
"min": 0,
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "rgba(50, 172, 45, 0.97)",
"value": 0
},
{
"color": "rgba(237, 129, 40, 0.89)",
"value": 80
},
{
"color": "rgba(245, 54, 54, 0.9)",
"value": 90
}
]
},
"unit": "percent"
},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 12,
"y": 1
},
"id": 16,
"options": {
"minVizHeight": 75,
"minVizWidth": 75,
"orientation": "auto",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showThresholdLabels": false,
"showThresholdMarkers": true,
"sizing": "auto"
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "clamp_min((1 - (node_memory_MemAvailable_bytes{server=\"${nodename}\", job=\"$job\"} / node_memory_MemTotal_bytes{server=\"${nodename}\", job=\"$job\"})) * 100, 0)",
"format": "time_series",
"instant": true,
"range": false,
"refId": "B",
"step": 240
}
],
"title": "RAM Used",
"type": "gauge"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "Used Root FS",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"max": 100,
"min": 0,
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "rgba(50, 172, 45, 0.97)",
"value": 0
},
{
"color": "rgba(237, 129, 40, 0.89)",
"value": 80
},
{
"color": "rgba(245, 54, 54, 0.9)",
"value": 90
}
]
},
"unit": "percent"
},
"overrides": []
},
"gridPos": {
"h": 4,
"w": 3,
"x": 15,
"y": 1
},
"id": 154,
"options": {
"minVizHeight": 75,
"minVizWidth": 75,
"orientation": "auto",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showThresholdLabels": false,
"showThresholdMarkers": true,
"sizing": "auto"
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "(\n (node_filesystem_size_bytes{server=\"${nodename}\", job=\"$job\", mountpoint=\"/\", fstype!=\"rootfs\"}\n - node_filesystem_avail_bytes{server=\"${nodename}\", job=\"$job\", mountpoint=\"/\", fstype!=\"rootfs\"})\n / node_filesystem_size_bytes{server=\"${nodename}\", job=\"$job\", mountpoint=\"/\", fstype!=\"rootfs\"}\n) * 100\n",
"format": "time_series",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "Root FS Used",
"type": "gauge"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": 0
}
]
},
"unit": "short"
},
"overrides": []
},
"gridPos": {
"h": 2,
"w": 2,
"x": 18,
"y": 1
},
"id": 14,
"maxDataPoints": 100,
"options": {
"colorMode": "none",
"graphMode": "none",
"justifyMode": "auto",
"orientation": "horizontal",
"percentChangeColorMode": "standard",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showPercentChange": false,
"textMode": "auto",
"wideLayout": true
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "count(count(node_cpu_seconds_total{server=\"${nodename}\",job=\"$job\"}) by (cpu))",
"instant": true,
"legendFormat": "__auto",
"range": false,
"refId": "A"
}
],
"title": "CPU Cores",
"type": "stat"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 0,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": 0
},
{
"color": "red",
"value": 80
}
]
},
"unit": "bytes"
},
"overrides": []
},
"gridPos": {
"h": 2,
"w": 2,
"x": 20,
"y": 1
},
"id": 75,
"maxDataPoints": 100,
"options": {
"colorMode": "none",
"graphMode": "none",
"justifyMode": "auto",
"orientation": "horizontal",
"percentChangeColorMode": "standard",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showPercentChange": false,
"textMode": "auto",
"wideLayout": true
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "node_memory_MemTotal_bytes{server=\"${nodename}\",job=\"$job\"}",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "RAM Total",
"type": "stat"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 0,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": 0
},
{
"color": "red",
"value": 80
}
]
},
"unit": "bytes"
},
"overrides": []
},
"gridPos": {
"h": 2,
"w": 2,
"x": 22,
"y": 1
},
"id": 18,
"maxDataPoints": 100,
"options": {
"colorMode": "none",
"graphMode": "none",
"justifyMode": "auto",
"orientation": "horizontal",
"percentChangeColorMode": "standard",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showPercentChange": false,
"textMode": "auto",
"wideLayout": true
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "node_memory_SwapTotal_bytes{server=\"${nodename}\",job=\"$job\"}",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "SWAP Total",
"type": "stat"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 1,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": 0
},
{
"color": "red",
"value": 80
}
]
},
"unit": "s"
},
"overrides": []
},
"gridPos": {
"h": 2,
"w": 4,
"x": 18,
"y": 3
},
"id": 15,
"maxDataPoints": 100,
"options": {
"colorMode": "none",
"graphMode": "none",
"justifyMode": "auto",
"orientation": "horizontal",
"percentChangeColorMode": "standard",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showPercentChange": false,
"textMode": "auto",
"wideLayout": true
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "node_time_seconds{server=\"${nodename}\",job=\"$job\"} - node_boot_time_seconds{server=\"${nodename}\",job=\"$job\"}",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "Uptime",
"type": "stat"
},
{
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"description": "",
"fieldConfig": {
"defaults": {
"color": {
"mode": "thresholds"
},
"decimals": 0,
"mappings": [
{
"options": {
"match": "null",
"result": {
"text": "N/A"
}
},
"type": "special"
}
],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "rgba(50, 172, 45, 0.97)",
"value": 0
},
{
"color": "rgba(237, 129, 40, 0.89)",
"value": 70
},
{
"color": "rgba(245, 54, 54, 0.9)",
"value": 90
}
]
},
"unit": "bytes"
},
"overrides": []
},
"gridPos": {
"h": 2,
"w": 2,
"x": 22,
"y": 3
},
"id": 23,
"maxDataPoints": 100,
"options": {
"colorMode": "none",
"graphMode": "none",
"justifyMode": "auto",
"orientation": "horizontal",
"percentChangeColorMode": "standard",
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"showPercentChange": false,
"textMode": "auto",
"wideLayout": true
},
"pluginVersion": "12.1.0",
"targets": [
{
"editorMode": "code",
"exemplar": false,
"expr": "node_filesystem_size_bytes{server=\"${nodename}\",job=\"$job\",mountpoint=\"/\",fstype!=\"rootfs\"}",
"format": "time_series",
"instant": true,
"range": false,
"refId": "A",
"step": 240
}
],
"title": "RootFS Total",
"type": "stat"
}
],
"preload": false,
"refresh": "1m",
"schemaVersion": 41,
"tags": [
"linux"
],
"templating": {
"list": [
{
"current": {
"text": "prometheus",
"value": "bezqqznn81wqof"
},
"includeAll": false,
"label": "Datasource",
"name": "ds_prometheus",
"options": [],
"query": "prometheus",
"refresh": 1,
"regex": "",
"type": "datasource"
},
{
"current": {
"text": "node_exporter",
"value": "node_exporter"
},
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"definition": "",
"includeAll": false,
"label": "Job",
"name": "job",
"options": [],
"query": {
"query": "label_values(node_uname_info, job)",
"refId": "Prometheus-job-Variable-Query"
},
"refresh": 1,
"regex": "",
"sort": 1,
"type": "query"
},
{
"current": {
"text": "All",
"value": "$__all"
},
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"definition": "label_values(node_uname_info{job=\"$job\"}, nodename)",
"includeAll": true,
"label": "Nodename",
"name": "nodename",
"options": [],
"query": {
"query": "label_values(node_uname_info{job=\"$job\"}, nodename)",
"refId": "Prometheus-nodename-Variable-Query"
},
"refresh": 1,
"regex": "",
"sort": 1,
"type": "query"
},
{
"current": {
"text": "",
"value": ""
},
"datasource": {
"type": "prometheus",
"uid": "${ds_prometheus}"
},
"definition": "label_values(node_uname_info{job=\"$job\", nodename=\"$nodename\"}, instance)",
"includeAll": false,
"label": "Instance",
"name": "node",
"options": [],
"query": {
"query": "label_values(node_uname_info{job=\"$job\", nodename=\"$nodename\"}, instance)",
"refId": "Prometheus-node-Variable-Query"
},
"refresh": 1,
"regex": "",
"sort": 1,
"type": "query"
}
]
},
"time": {
"from": "now-15m",
"to": "now"
},
"timepicker": {
"nowDelay": "1m"
},
"timezone": "browser",
"title": "Infrastructure",
"uid": "rYdddlPWkd"
}