mirror of
https://github.com/Alkatrazz24/Funk-lab.git
synced 2026-07-08 10:04:42 +02:00
ALI YESILKAYA
470c69881c
Trois instances llama-server CPU manuelles (hors Ansible) tournaient sur gpu-01
sans servir : déjà orphelines de LiteLLM (alias absents de config.yaml, d'où le
bruit "Invalid model name"). Suppression complète :
gpu-01 (opérationnel, déjà fait) :
- systemctl disable --now + rm des units llama-server-{system,monitor,dev}
- ports firewall 1235/1236/1237 fermés
Repo :
- monitoring : scrape jobs (values.yaml) + alertes LlamaServer{System,Monitor}Down
(alerts-ai.yaml) retirés → plus de "target down"/alertes fantômes ; panneaux de
statut dashboard-infrastructure recyclés (system→llama-embed, monitor supprimé).
- docs admin/ + hermes-skills (souls funk-ai/brain, SKILL agent-delegation) mises à
jour : inventaire modèles, profils Hermès system/monitor sans backend (à repointer
sur qwen3-8b), litellm, ops, monitoring.
Les profils Hermès system/monitor existent toujours mais sans backend dédié.
Dashboards Grafana dashboard-ai (panneaux graphiques CPU) laissés tels quels
(cosmétique, séries vides, aucune alerte).
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
1511 lines
57 KiB
YAML
1511 lines
57 KiB
YAML
apiVersion: v1
|
|
kind: ConfigMap
|
|
metadata:
|
|
name: grafana-dashboard-infrastructure
|
|
namespace: monitoring
|
|
labels:
|
|
grafana_dashboard: "1"
|
|
data:
|
|
infrastructure.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": "Compute nodes (compute-01 / 02 / 03)",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 1},
|
|
"id": 2,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "CPU % (avg compute)",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "avg(100 - (avg by (instance) (irate(node_cpu_seconds_total{mode=\"idle\",job=\"node-exporter\"}[5m])) * 100))",
|
|
"legendFormat": "CPU avg",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 1},
|
|
"id": 3,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "RAM % (avg compute)",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "avg(100 - (node_memory_MemAvailable_bytes{job=\"node-exporter\"} / node_memory_MemTotal_bytes{job=\"node-exporter\"} * 100))",
|
|
"legendFormat": "RAM avg",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 2},
|
|
{"color": "red", "value": 4}
|
|
]
|
|
},
|
|
"unit": "short",
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 1},
|
|
"id": 4,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Load1 (avg compute)",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "avg(node_load1{job=\"node-exporter\"})",
|
|
"legendFormat": "Load1",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "yellow", "value": null},
|
|
{"color": "green", "value": 43200},
|
|
{"color": "blue", "value": 604800}
|
|
]
|
|
},
|
|
"unit": "dtdurations",
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 6, "x": 12, "y": 1},
|
|
"id": 40,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["min"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Uptime min — compute",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "time() - node_boot_time_seconds{job=\"node-exporter\"}",
|
|
"legendFormat": "{{instance}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 60},
|
|
{"color": "red", "value": 85}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 6, "x": 18, "y": 1},
|
|
"id": 41,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Disk /var avg % — compute",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "avg(100 - (node_filesystem_avail_bytes{job=\"node-exporter\",mountpoint=\"/var\",fstype=\"xfs\"} / node_filesystem_size_bytes{job=\"node-exporter\",mountpoint=\"/var\",fstype=\"xfs\"} * 100))",
|
|
"legendFormat": "/var avg",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 10,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 8, "x": 0, "y": 5},
|
|
"id": 11,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "CPU % — compute-01/02/03",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (avg by (instance) (irate(node_cpu_seconds_total{mode=\"idle\",job=\"node-exporter\"}[5m])) * 100)",
|
|
"legendFormat": "{{instance}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 10,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 8, "x": 8, "y": 5},
|
|
"id": 12,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "RAM % — compute-01/02/03",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (node_memory_MemAvailable_bytes{job=\"node-exporter\"} / node_memory_MemTotal_bytes{job=\"node-exporter\"} * 100)",
|
|
"legendFormat": "{{instance}}",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 5,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "Bps",
|
|
"min": 0
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 8, "x": 16, "y": 5},
|
|
"id": 13,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "Network I/O — compute nodes",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum by (instance) (irate(node_network_receive_bytes_total{job=\"node-exporter\",device!~\"lo|veth.*|flannel.*|cni.*\"}[5m]))",
|
|
"legendFormat": "rx {{instance}}",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum by (instance) (irate(node_network_transmit_bytes_total{job=\"node-exporter\",device!~\"lo|veth.*|flannel.*|cni.*\"}[5m]))",
|
|
"legendFormat": "tx {{instance}}",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 13},
|
|
"id": 20,
|
|
"title": "storage-01 (gateway — RAID5 — NFS)",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"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": "none"
|
|
},
|
|
"title": "CPU % — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (avg(irate(node_cpu_seconds_total{mode=\"idle\",job=\"storage-01\"}[5m])) * 100)",
|
|
"legendFormat": "CPU",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"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": "none"
|
|
},
|
|
"title": "RAM % — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (node_memory_MemAvailable_bytes{job=\"storage-01\"} / node_memory_MemTotal_bytes{job=\"storage-01\"} * 100)",
|
|
"legendFormat": "RAM",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"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": "none"
|
|
},
|
|
"title": "Disk /srv/data % — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (node_filesystem_avail_bytes{job=\"storage-01\",mountpoint=\"/srv/data\",fstype!=\"tmpfs\"} / node_filesystem_size_bytes{job=\"storage-01\",mountpoint=\"/srv/data\",fstype!=\"tmpfs\"} * 100)",
|
|
"legendFormat": "/srv/data",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [
|
|
{
|
|
"type": "value",
|
|
"options": {
|
|
"1": {"text": "ACTIF", "color": "green", "index": 0},
|
|
"0": {"text": "DÉGRADÉ", "color": "red", "index": 1}
|
|
}
|
|
}
|
|
]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 3, "x": 12, "y": 14},
|
|
"id": 42,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "RAID5 état",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_md_state{job=\"storage-01\",device=\"md0\",state=\"active\"}",
|
|
"legendFormat": "md0",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "red", "value": 1}
|
|
]
|
|
},
|
|
"unit": "none",
|
|
"mappings": [],
|
|
"noValue": "0"
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 3, "x": 15, "y": 14},
|
|
"id": 43,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "RAID5 disques HS",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_md_disks{job=\"storage-01\",device=\"md0\",state=\"failed\"}",
|
|
"legendFormat": "failed",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "yellow", "value": null},
|
|
{"color": "green", "value": 43200},
|
|
{"color": "blue", "value": 604800}
|
|
]
|
|
},
|
|
"unit": "dtdurations",
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 3, "x": 18, "y": 14},
|
|
"id": 44,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Uptime — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "time() - node_boot_time_seconds{job=\"storage-01\"}",
|
|
"legendFormat": "uptime",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 60},
|
|
{"color": "red", "value": 80}
|
|
]
|
|
},
|
|
"unit": "celsius",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 3, "x": 21, "y": 14},
|
|
"id": 45,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Temp CPU — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"storage-01\",chip=\"pci0000:00_0000:00:18_3\",sensor=\"temp1\"}",
|
|
"legendFormat": "CPU",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 10,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 0, "y": 18},
|
|
"id": 24,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "CPU — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (avg(irate(node_cpu_seconds_total{mode=\"idle\",job=\"storage-01\"}[5m])) * 100)",
|
|
"legendFormat": "CPU %",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 10,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 6, "y": 18},
|
|
"id": 25,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "RAM — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (node_memory_MemAvailable_bytes{job=\"storage-01\"} / node_memory_MemTotal_bytes{job=\"storage-01\"} * 100)",
|
|
"legendFormat": "RAM %",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 5,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "Bps",
|
|
"min": 0
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 12, "y": 18},
|
|
"id": 26,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "Disk I/O — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum(irate(node_disk_read_bytes_total{job=\"storage-01\"}[5m]))",
|
|
"legendFormat": "read",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum(irate(node_disk_written_bytes_total{job=\"storage-01\"}[5m]))",
|
|
"legendFormat": "write",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 5,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "Bps",
|
|
"min": 0
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 18, "y": 18},
|
|
"id": 27,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "Network I/O — storage-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum(irate(node_network_receive_bytes_total{job=\"storage-01\",device!~\"lo|veth.*\"}[5m]))",
|
|
"legendFormat": "rx",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "sum(irate(node_network_transmit_bytes_total{job=\"storage-01\",device!~\"lo|veth.*\"}[5m]))",
|
|
"legendFormat": "tx",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 26},
|
|
"id": 30,
|
|
"title": "gpu-01 (RX 6700XT — ROCm — llama-server)",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 27},
|
|
"id": 31,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "CPU % — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (avg(irate(node_cpu_seconds_total{mode=\"idle\",job=\"gpu-01-node\"}[5m])) * 100)",
|
|
"legendFormat": "CPU",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 27},
|
|
"id": 32,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "RAM % — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "100 - (node_memory_MemAvailable_bytes{job=\"gpu-01-node\"} / node_memory_MemTotal_bytes{job=\"gpu-01-node\"} * 100)",
|
|
"legendFormat": "RAM",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "yellow", "value": null},
|
|
{"color": "green", "value": 43200},
|
|
{"color": "blue", "value": 604800}
|
|
]
|
|
},
|
|
"unit": "dtdurations",
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 5, "x": 8, "y": 27},
|
|
"id": 46,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Uptime — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "time() - node_boot_time_seconds{job=\"gpu-01-node\"}",
|
|
"legendFormat": "uptime",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "celsius",
|
|
"min": 0,
|
|
"max": 110,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 5, "x": 13, "y": 27},
|
|
"id": 47,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Temp CPU Tdie — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"pci0000:00_0000:00:18_3\",sensor=\"temp3\"}",
|
|
"legendFormat": "Tdie",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 90}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 4, "w": 6, "x": 18, "y": 27},
|
|
"id": 48,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"textMode": "auto",
|
|
"colorMode": "background",
|
|
"graphMode": "none"
|
|
},
|
|
"title": "Swap % — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "(1 - node_memory_SwapFree_bytes{job=\"gpu-01-node\"} / node_memory_SwapTotal_bytes{job=\"gpu-01-node\"}) * 100",
|
|
"legendFormat": "Swap",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {
|
|
"mode": "absolute",
|
|
"steps": [
|
|
{"color": "green", "value": null},
|
|
{"color": "yellow", "value": 70},
|
|
{"color": "red", "value": 85}
|
|
]
|
|
},
|
|
"unit": "celsius",
|
|
"min": 0,
|
|
"max": 110,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 0, "y": 31},
|
|
"id": 33,
|
|
"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": 80},
|
|
{"color": "red", "value": 92}
|
|
]
|
|
},
|
|
"unit": "percent",
|
|
"min": 0,
|
|
"max": 100,
|
|
"mappings": []
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 6, "x": 6, "y": 31},
|
|
"id": 34,
|
|
"options": {
|
|
"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false},
|
|
"orientation": "auto",
|
|
"showThresholdLabels": false,
|
|
"showThresholdMarkers": true
|
|
},
|
|
"title": "VRAM Usage %",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "rocm_vram_used_bytes{job=\"gpu-01-node\",gpu=\"0\"} / rocm_vram_total_bytes{job=\"gpu-01-node\",gpu=\"0\"} * 100",
|
|
"legendFormat": "VRAM %",
|
|
"refId": "A"
|
|
}
|
|
],
|
|
"type": "gauge"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${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": 31},
|
|
"id": 35,
|
|
"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 %"},
|
|
"properties": [{"id": "unit", "value": "percent"}, {"id": "custom.axisPlacement", "value": "right"}, {"id": "max", "value": 100}]
|
|
}
|
|
]
|
|
},
|
|
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 39},
|
|
"id": 36,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "GPU Temp + VRAM % — gpu-01",
|
|
"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\"} / rocm_vram_total_bytes{job=\"gpu-01-node\",gpu=\"0\"} * 100",
|
|
"legendFormat": "VRAM %",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "celsius",
|
|
"min": 0,
|
|
"max": 110
|
|
},
|
|
"overrides": [
|
|
{
|
|
"matcher": {"id": "byName", "options": "edge"},
|
|
"properties": [{"id": "color", "value": {"fixedColor": "blue", "mode": "fixed"}}]
|
|
},
|
|
{
|
|
"matcher": {"id": "byName", "options": "hotspot"},
|
|
"properties": [{"id": "color", "value": {"fixedColor": "orange", "mode": "fixed"}}]
|
|
},
|
|
{
|
|
"matcher": {"id": "byName", "options": "VRAM"},
|
|
"properties": [{"id": "color", "value": {"fixedColor": "purple", "mode": "fixed"}}]
|
|
}
|
|
]
|
|
},
|
|
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 47},
|
|
"id": 49,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "GPU Températures (edge / hotspot / VRAM) — gpu-01",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"0000:09:00_0_0000:0a:00_0\",sensor=\"temp1\"}",
|
|
"legendFormat": "edge",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"0000:09:00_0_0000:0a:00_0\",sensor=\"temp2\"}",
|
|
"legendFormat": "hotspot",
|
|
"refId": "B"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"0000:09:00_0_0000:0a:00_0\",sensor=\"temp3\"}",
|
|
"legendFormat": "VRAM",
|
|
"refId": "C"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {
|
|
"lineWidth": 1,
|
|
"fillOpacity": 8,
|
|
"drawStyle": "line",
|
|
"spanNulls": false
|
|
},
|
|
"unit": "celsius",
|
|
"min": 0,
|
|
"max": 110
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 47},
|
|
"id": 50,
|
|
"options": {
|
|
"tooltip": {"mode": "multi"},
|
|
"legend": {"displayMode": "list", "placement": "bottom"}
|
|
},
|
|
"title": "CPU Températures — gpu-01 (AMD Tdie)",
|
|
"targets": [
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"pci0000:00_0000:00:18_3\",sensor=\"temp3\"}",
|
|
"legendFormat": "Tdie",
|
|
"refId": "A"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"expr": "node_hwmon_temp_celsius{job=\"gpu-01-node\",chip=\"pci0000:00_0000:00:18_3\",sensor=\"temp1\"}",
|
|
"legendFormat": "Tctl",
|
|
"refId": "B"
|
|
}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 55},
|
|
"id": 51,
|
|
"title": "Services — état systemd",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 56},
|
|
"id": 70,
|
|
"title": "storage-01",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "dnsmasq", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 0, "y": 57},
|
|
"id": 52,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "DNS (dnsmasq)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"dnsmasq.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "nfs-server", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 3, "y": 57},
|
|
"id": 53,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "NFS server",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"nfs-server.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "postgresql", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 6, "y": 57},
|
|
"id": 54,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "PostgreSQL",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"postgresql-16.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "qdrant", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 9, "y": 57},
|
|
"id": 55,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "Qdrant",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"qdrant.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "litellm", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 12, "y": 57},
|
|
"id": 56,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "LiteLLM",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"litellm.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "hermes-agent", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 15, "y": 57},
|
|
"id": 57,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "Hermes Agent",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"hermes-agent.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "nftables", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 18, "y": 57},
|
|
"id": 58,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "nftables (firewall)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"nftables.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "chronyd", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 3, "x": 21, "y": 57},
|
|
"id": 59,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "NTP (chronyd)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"storage-01\",name=\"chronyd.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"collapsed": false,
|
|
"gridPos": {"h": 1, "w": 24, "x": 0, "y": 60},
|
|
"id": 71,
|
|
"title": "gpu-01",
|
|
"type": "row"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "llama-server GPU", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 6, "x": 0, "y": 61},
|
|
"id": 60,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "llama-server GPU (:1234)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"gpu-01-node\",name=\"llama-server.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "llama-embed nomic", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 6, "x": 6, "y": 61},
|
|
"id": 61,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "llama-embed nomic (:1238)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"gpu-01-node\",name=\"llama-embed.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "thresholds"},
|
|
"thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]},
|
|
"unit": "none",
|
|
"mappings": [{"type": "value", "options": {"1": {"text": "chronyd", "color": "green", "index": 0}, "0": {"text": "ARRÊTÉ", "color": "red", "index": 1}}}]
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 3, "w": 6, "x": 18, "y": 61},
|
|
"id": 63,
|
|
"options": {"reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto", "colorMode": "background", "graphMode": "none"},
|
|
"title": "NTP (chronyd)",
|
|
"targets": [{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "node_systemd_unit_state{job=\"gpu-01-node\",name=\"chronyd.service\",state=\"active\"}", "legendFormat": "", "refId": "A"}],
|
|
"type": "stat"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {"lineWidth": 1, "fillOpacity": 8, "drawStyle": "line", "spanNulls": false},
|
|
"unit": "Bps",
|
|
"min": 0
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 7, "w": 12, "x": 0, "y": 64},
|
|
"id": 64,
|
|
"options": {"tooltip": {"mode": "multi"}, "legend": {"displayMode": "list", "placement": "bottom"}},
|
|
"title": "NFS I/O — storage-01 (débit disque NFS)",
|
|
"targets": [
|
|
{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "irate(node_nfsd_disk_bytes_read_total{job=\"storage-01\"}[5m])", "legendFormat": "read", "refId": "A"},
|
|
{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "irate(node_nfsd_disk_bytes_written_total{job=\"storage-01\"}[5m])", "legendFormat": "write", "refId": "B"}
|
|
],
|
|
"type": "timeseries"
|
|
},
|
|
{
|
|
"datasource": {"type": "prometheus", "uid": "${datasource}"},
|
|
"fieldConfig": {
|
|
"defaults": {
|
|
"color": {"mode": "palette-classic"},
|
|
"custom": {"lineWidth": 1, "fillOpacity": 8, "drawStyle": "line", "spanNulls": false},
|
|
"unit": "reqps",
|
|
"min": 0
|
|
},
|
|
"overrides": []
|
|
},
|
|
"gridPos": {"h": 7, "w": 12, "x": 12, "y": 64},
|
|
"id": 65,
|
|
"options": {"tooltip": {"mode": "multi"}, "legend": {"displayMode": "list", "placement": "bottom"}},
|
|
"title": "NFS requêtes/s par type — storage-01",
|
|
"targets": [
|
|
{"datasource": {"type": "prometheus", "uid": "${datasource}"}, "expr": "irate(node_nfsd_requests_total{job=\"storage-01\",method=~\"read|write|getattr|lookup\"}[5m])", "legendFormat": "{{method}}", "refId": "A"}
|
|
],
|
|
"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 — Infrastructure",
|
|
"uid": "funk-inf1",
|
|
"version": 1
|
|
}
|