Funk-lab/k8s/infra/monitoring/dashboards/dashboard-ai.yaml
alkatrazz 18da44a89f feat: monitoring — dashboards Grafana + alertes Prometheus
Dashboards (ConfigMaps, auto-importés par sidecar Grafana) :
- dashboard-kubernetes : cluster overview, pods, PVCs, restarts
- dashboard-infrastructure : compute/storage-01/gpu-01 CPU/RAM/disque/réseau + GPU ROCm
- dashboard-ai : llama-server GPU/CPU tokens/s, requêtes, VRAM, température GPU

Alertes (PrometheusRule) :
- alerts-node : CPU>90%, RAM>90%, disque>80%, storage-01/gpu-01 down, load élevé
- alerts-k8s : NodeNotReady, PodCrashLoop, PodPending, DeploymentUnavailable, PVCNotBound, ArgoCDOutOfSync
- alerts-ai : llama-server GPU/system/monitor down, GPU temp>80°C, VRAM>95%, queue élevée

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-13 22:47:49 +02:00

753 lines
25 KiB
YAML

apiVersion: v1
kind: ConfigMap
metadata:
name: grafana-dashboard-ai
namespace: monitoring
labels:
grafana_dashboard: "1"
data:
ai.json: |
{
"__inputs": [],
"__requires": [],
"annotations": {"list": []},
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 1,
"id": null,
"links": [],
"panels": [
{
"collapsed": false,
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 0},
"id": 1,
"title": "GPU — Qwen3-8B (llama-server GPU, port 1234)",
"type": "row"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "blue", "value": null}
]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 1},
"id": 2,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Génération tokens/s (GPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:predicted_tokens_seconds{job=\"llama-server-gpu\"}",
"legendFormat": "gen t/s",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "blue", "value": null}
]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 1},
"id": 3,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Prompt tokens/s (GPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:prompt_tokens_seconds{job=\"llama-server-gpu\"}",
"legendFormat": "prompt t/s",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 2},
{"color": "red", "value": 5}
]
},
"unit": "short",
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 1},
"id": 4,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "none"
},
"title": "Requêtes en cours (GPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:requests_processing{job=\"llama-server-gpu\"}",
"legendFormat": "processing",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 1},
{"color": "red", "value": 3}
]
},
"unit": "short",
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 1},
"id": 5,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "none"
},
"title": "Requêtes en attente (GPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:requests_deferred{job=\"llama-server-gpu\"}",
"legendFormat": "deferred",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "palette-classic"},
"custom": {
"lineWidth": 2,
"fillOpacity": 10,
"drawStyle": "line",
"spanNulls": false
},
"unit": "short",
"min": 0,
"decimals": 1
},
"overrides": []
},
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 5},
"id": 11,
"options": {
"tooltip": {"mode": "multi"},
"legend": {"displayMode": "list", "placement": "bottom"}
},
"title": "Tokens/s — génération + prompt (GPU Qwen3-8B)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:predicted_tokens_seconds{job=\"llama-server-gpu\"}",
"legendFormat": "génération t/s",
"refId": "A"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:prompt_tokens_seconds{job=\"llama-server-gpu\"}",
"legendFormat": "prompt t/s",
"refId": "B"
}
],
"type": "timeseries"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "palette-classic"},
"custom": {
"lineWidth": 1,
"fillOpacity": 10,
"drawStyle": "line",
"spanNulls": false
},
"unit": "short",
"min": 0
},
"overrides": []
},
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 5},
"id": 12,
"options": {
"tooltip": {"mode": "multi"},
"legend": {"displayMode": "list", "placement": "bottom"}
},
"title": "Tokens prédits (rate 5m) — GPU",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rate(llamacpp:tokens_predicted_total{job=\"llama-server-gpu\"}[5m])",
"legendFormat": "tokens/s (5m rate)",
"refId": "A"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rate(llamacpp:prompt_tokens_total{job=\"llama-server-gpu\"}[5m])",
"legendFormat": "prompt tokens/s (5m rate)",
"refId": "B"
}
],
"type": "timeseries"
},
{
"collapsed": false,
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 13},
"id": 20,
"title": "CPU — system + monitor (Qwen3-1.7B, ports 1236 / 1237)",
"type": "row"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [{"color": "blue", "value": null}]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 14},
"id": 21,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Génération t/s — system (CPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:predicted_tokens_seconds{job=\"llama-server-system\"}",
"legendFormat": "gen t/s system",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [{"color": "blue", "value": null}]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 14},
"id": 22,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Génération t/s — monitor (CPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:predicted_tokens_seconds{job=\"llama-server-monitor\"}",
"legendFormat": "gen t/s monitor",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [{"color": "blue", "value": null}]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 14},
"id": 23,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Prompt t/s — system (CPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:prompt_tokens_seconds{job=\"llama-server-system\"}",
"legendFormat": "prompt t/s system",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [{"color": "blue", "value": null}]
},
"unit": "short",
"decimals": 1,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 14},
"id": 24,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "area"
},
"title": "Prompt t/s — monitor (CPU)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:prompt_tokens_seconds{job=\"llama-server-monitor\"}",
"legendFormat": "prompt t/s monitor",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 2},
{"color": "red", "value": 5}
]
},
"unit": "short",
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 16, "y": 14},
"id": 25,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "none"
},
"title": "Requêtes en cours — system + monitor",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:requests_processing{job=~\"llama-server-system|llama-server-monitor\"}",
"legendFormat": "{{job}}",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 1},
{"color": "red", "value": 3}
]
},
"unit": "short",
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 4, "w": 4, "x": 20, "y": 14},
"id": 26,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"textMode": "auto",
"colorMode": "background",
"graphMode": "none"
},
"title": "Requêtes en attente — system + monitor",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:requests_deferred{job=~\"llama-server-system|llama-server-monitor\"}",
"legendFormat": "{{job}}",
"refId": "A"
}
],
"type": "stat"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "palette-classic"},
"custom": {
"lineWidth": 2,
"fillOpacity": 10,
"drawStyle": "line",
"spanNulls": false
},
"unit": "short",
"min": 0,
"decimals": 1
},
"overrides": []
},
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 18},
"id": 27,
"options": {
"tooltip": {"mode": "multi"},
"legend": {"displayMode": "list", "placement": "bottom"}
},
"title": "Tokens/s — system + monitor (CPU Qwen3-1.7B)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:predicted_tokens_seconds{job=~\"llama-server-system|llama-server-monitor\"}",
"legendFormat": "gen {{job}}",
"refId": "A"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "llamacpp:prompt_tokens_seconds{job=~\"llama-server-system|llama-server-monitor\"}",
"legendFormat": "prompt {{job}}",
"refId": "B"
}
],
"type": "timeseries"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "palette-classic"},
"custom": {
"lineWidth": 1,
"fillOpacity": 10,
"drawStyle": "line",
"spanNulls": false
},
"unit": "short",
"min": 0
},
"overrides": []
},
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 18},
"id": 28,
"options": {
"tooltip": {"mode": "multi"},
"legend": {"displayMode": "list", "placement": "bottom"}
},
"title": "Tokens prédits (rate 5m) — system + monitor",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rate(llamacpp:tokens_predicted_total{job=~\"llama-server-system|llama-server-monitor\"}[5m])",
"legendFormat": "tokens/s {{job}}",
"refId": "A"
}
],
"type": "timeseries"
},
{
"collapsed": false,
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 26},
"id": 40,
"title": "GPU Health (RX 6700XT — gfx1031)",
"type": "row"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 70},
{"color": "red", "value": 85}
]
},
"unit": "celsius",
"min": 0,
"max": 110,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 8, "w": 6, "x": 0, "y": 27},
"id": 41,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"showThresholdLabels": false,
"showThresholdMarkers": true
},
"title": "GPU Temperature (°C)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rocm_gpu_temperature_celsius{job=\"gpu-01-node\",gpu=\"0\"}",
"legendFormat": "GPU Temp",
"refId": "A"
}
],
"type": "gauge"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 10737418240},
{"color": "red", "value": 11877376000}
]
},
"unit": "bytes",
"min": 0,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 8, "w": 6, "x": 6, "y": 27},
"id": 42,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"showThresholdLabels": false,
"showThresholdMarkers": true
},
"title": "VRAM Usage (GB used / 12.9 GB total)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rocm_vram_used_bytes{job=\"gpu-01-node\",gpu=\"0\"}",
"legendFormat": "VRAM used",
"refId": "A"
}
],
"type": "gauge"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "thresholds"},
"thresholds": {
"mode": "absolute",
"steps": [
{"color": "green", "value": null},
{"color": "yellow", "value": 70},
{"color": "red", "value": 90}
]
},
"unit": "percent",
"min": 0,
"max": 100,
"mappings": []
},
"overrides": []
},
"gridPos": {"h": 8, "w": 6, "x": 12, "y": 27},
"id": 43,
"options": {
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
"orientation": "auto",
"showThresholdLabels": false,
"showThresholdMarkers": true
},
"title": "GPU Utilization %",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rocm_gpu_utilization_percent{job=\"gpu-01-node\",gpu=\"0\"}",
"legendFormat": "GPU Util %",
"refId": "A"
}
],
"type": "gauge"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"fieldConfig": {
"defaults": {
"color": {"mode": "palette-classic"},
"custom": {
"lineWidth": 1,
"fillOpacity": 10,
"drawStyle": "line",
"spanNulls": false
},
"min": 0
},
"overrides": [
{
"matcher": {"id": "byName", "options": "GPU Temp (°C)"},
"properties": [
{"id": "unit", "value": "celsius"},
{"id": "custom.axisPlacement", "value": "left"}
]
},
{
"matcher": {"id": "byName", "options": "VRAM used (GB)"},
"properties": [
{"id": "unit", "value": "bytes"},
{"id": "custom.axisPlacement", "value": "right"}
]
}
]
},
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 35},
"id": 44,
"options": {
"tooltip": {"mode": "multi"},
"legend": {"displayMode": "list", "placement": "bottom"}
},
"title": "GPU Temp + VRAM usage — gpu-01 (gfx1031)",
"targets": [
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rocm_gpu_temperature_celsius{job=\"gpu-01-node\",gpu=\"0\"}",
"legendFormat": "GPU Temp (°C)",
"refId": "A"
},
{
"datasource": {"type": "prometheus", "uid": "${datasource}"},
"expr": "rocm_vram_used_bytes{job=\"gpu-01-node\",gpu=\"0\"}",
"legendFormat": "VRAM used (GB)",
"refId": "B"
}
],
"type": "timeseries"
}
],
"refresh": "30s",
"schemaVersion": 38,
"tags": ["funk"],
"templating": {
"list": [
{
"current": {},
"hide": 0,
"includeAll": false,
"label": "Datasource",
"name": "datasource",
"options": [],
"query": "prometheus",
"refresh": 1,
"type": "datasource"
}
]
},
"time": {"from": "now-3h", "to": "now"},
"timepicker": {},
"timezone": "Europe/Paris",
"title": "Funk — IA / Hermes",
"uid": "funk-ai01",
"version": 1
}