groot
e0bf42bf32
mega-ci / static-release-gates (push) Failing after 11s
release-tag / release-image (push) Successful in 6m46s
mega-ci / go-quality (services/control) (push) Successful in 10m7s
mega-ci / go-quality (platform/neuroforge) (push) Successful in 10m20s
mega-ci / go-quality (services/agent) (push) Successful in 11m1s
mega-ci / go-quality (services/knowledge) (push) Successful in 11m13s
mega-ci / docker-build (push) Has been skipped
1488 lines
31 KiB
JSON
1488 lines
31 KiB
JSON
{
|
|
"id": null,
|
|
"uid": "neuroforge-master-subagents",
|
|
"title": "NeuroForge Master / Subagents / Knowledge Graph",
|
|
"tags": [
|
|
"neuroforge",
|
|
"orchestrator",
|
|
"knowledge-graph"
|
|
],
|
|
"timezone": "browser",
|
|
"schemaVersion": 41,
|
|
"version": 2,
|
|
"refresh": "15s",
|
|
"time": {
|
|
"from": "now-6h",
|
|
"to": "now"
|
|
},
|
|
"timepicker": {},
|
|
"templating": {
|
|
"list": []
|
|
},
|
|
"annotations": {
|
|
"list": []
|
|
},
|
|
"panels": [
|
|
{
|
|
"id": 1,
|
|
"type": "stat",
|
|
"title": "Master Up",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_up",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 2,
|
|
"type": "stat",
|
|
"title": "Memories",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 4,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_memories",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 3,
|
|
"type": "stat",
|
|
"title": "Synapses",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 8,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_synapses",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 4,
|
|
"type": "stat",
|
|
"title": "Online CPU Workers",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "sum(neuroforge_workers{resource=\"cpu\",status=\"online\"})",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 5,
|
|
"type": "stat",
|
|
"title": "Online GPU Workers",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 16,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "sum(neuroforge_workers{resource=\"gpu\",status=\"online\"})",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 6,
|
|
"type": "stat",
|
|
"title": "Queued Jobs",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 20,
|
|
"y": 0,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_jobs{status=\"queued\"}",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 7,
|
|
"type": "stat",
|
|
"title": "Linked Memories",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_memories{state=\"linked\"}",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 8,
|
|
"type": "stat",
|
|
"title": "Isolated Memories",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 4,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_memories{state=\"isolated\"}",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 9,
|
|
"type": "stat",
|
|
"title": "Multi-linked",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 8,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_memories{state=\"multi_linked\"}",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 10,
|
|
"type": "stat",
|
|
"title": "Average Degree",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_average_degree",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 11,
|
|
"type": "stat",
|
|
"title": "Connected Components",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 16,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_connected_components",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 12,
|
|
"type": "stat",
|
|
"title": "Largest Component",
|
|
"description": "",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 20,
|
|
"y": 4,
|
|
"w": 4,
|
|
"h": 4
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_largest_component",
|
|
"instant": true,
|
|
"range": false
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"mappings": [],
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"reduceOptions": {
|
|
"values": false,
|
|
"calcs": [
|
|
"lastNotNull"
|
|
],
|
|
"fields": ""
|
|
},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "value",
|
|
"graphMode": "area",
|
|
"justifyMode": "auto"
|
|
}
|
|
},
|
|
{
|
|
"id": 13,
|
|
"type": "timeseries",
|
|
"title": "Orchestrator Jobs by Status",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 8,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_jobs",
|
|
"legendFormat": "{{status}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 14,
|
|
"type": "timeseries",
|
|
"title": "Worker Inflight",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 8,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_worker_inflight",
|
|
"legendFormat": "{{worker}} / {{resource}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 15,
|
|
"type": "timeseries",
|
|
"title": "Worker Capacity",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 16,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_worker_capacity",
|
|
"legendFormat": "{{worker}} / {{resource}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 16,
|
|
"type": "timeseries",
|
|
"title": "Graph Linkage",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 16,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_graph_memories",
|
|
"legendFormat": "{{state}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 17,
|
|
"type": "timeseries",
|
|
"title": "Vector Index Population",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 24,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_hnsw_nodes",
|
|
"legendFormat": "HNSW",
|
|
"range": true
|
|
},
|
|
{
|
|
"refId": "B",
|
|
"expr": "neuroforge_disk_pq_items",
|
|
"legendFormat": "Disk PQ",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "short",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 18,
|
|
"type": "timeseries",
|
|
"title": "HTTP Request Rate",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 24,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "sum by (code) (rate(neuroforge_http_requests_total[5m]))",
|
|
"legendFormat": "HTTP {{code}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "reqps",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 19,
|
|
"type": "timeseries",
|
|
"title": "HTTP P95 Latency",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 32,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "histogram_quantile(0.95, sum by (le) (rate(neuroforge_http_request_duration_seconds_bucket[5m])))",
|
|
"legendFormat": "p95",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "s",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 20,
|
|
"type": "timeseries",
|
|
"title": "Master Heap",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 32,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "neuroforge_runtime_heap_alloc_bytes",
|
|
"legendFormat": "heap alloc",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "bytes",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 21,
|
|
"type": "timeseries",
|
|
"title": "Master / CPU / GPU Host CPU",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 40,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "100 - (avg by (job) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m])) * 100)",
|
|
"legendFormat": "{{job}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 22,
|
|
"type": "timeseries",
|
|
"title": "Host Memory Usage",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 12,
|
|
"y": 40,
|
|
"w": 12,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "100 * (1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes))",
|
|
"legendFormat": "{{job}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 23,
|
|
"type": "timeseries",
|
|
"title": "GPU Utilization",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 48,
|
|
"w": 8,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "avg by (gpu) (DCGM_FI_DEV_GPU_UTIL)",
|
|
"legendFormat": "GPU {{gpu}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "percent",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 24,
|
|
"type": "timeseries",
|
|
"title": "GPU Framebuffer Used",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 8,
|
|
"y": 48,
|
|
"w": 8,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "DCGM_FI_DEV_FB_USED",
|
|
"legendFormat": "GPU {{gpu}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "MiB",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 25,
|
|
"type": "timeseries",
|
|
"title": "GPU Power",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 16,
|
|
"y": 48,
|
|
"w": 8,
|
|
"h": 8
|
|
},
|
|
"targets": [
|
|
{
|
|
"refId": "A",
|
|
"expr": "DCGM_FI_DEV_POWER_USAGE",
|
|
"legendFormat": "GPU {{gpu}}",
|
|
"range": true
|
|
}
|
|
],
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "watt",
|
|
"custom": {
|
|
"drawStyle": "line",
|
|
"lineInterpolation": "linear",
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"showPoints": "auto",
|
|
"spanNulls": false
|
|
},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{
|
|
"color": "green",
|
|
"value": null
|
|
}
|
|
]
|
|
}
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"calcs": [
|
|
"lastNotNull"
|
|
]
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "desc"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"id": 26,
|
|
"title": "Durable Job Payload Bytes",
|
|
"type": "timeseries",
|
|
"datasource": {
|
|
"type": "prometheus",
|
|
"uid": "prometheus"
|
|
},
|
|
"gridPos": {
|
|
"x": 0,
|
|
"y": 56,
|
|
"w": 24,
|
|
"h": 8
|
|
},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"unit": "bytes"
|
|
},
|
|
"overrides": []
|
|
},
|
|
"options": {
|
|
"legend": {
|
|
"displayMode": "list",
|
|
"placement": "bottom",
|
|
"showLegend": true
|
|
},
|
|
"tooltip": {
|
|
"mode": "multi",
|
|
"sort": "none"
|
|
}
|
|
},
|
|
"targets": [
|
|
{
|
|
"expr": "neuroforge_job_payload_bytes",
|
|
"legendFormat": "{{state}}",
|
|
"refId": "A"
|
|
}
|
|
]
|
|
}
|
|
]
|
|
}
|