feat(ai-stack): add DCGM exporter, Prometheus scraper, and Grafana AI Stack intelligence dashboard
This commit is contained in:
27
argocd/aerodrome/prometheus-app.yaml
Normal file
27
argocd/aerodrome/prometheus-app.yaml
Normal file
@@ -0,0 +1,27 @@
|
||||
apiVersion: argoproj.io/v1alpha1
|
||||
kind: Application
|
||||
metadata:
|
||||
name: aerodrome-prometheus
|
||||
namespace: infrastructure
|
||||
labels:
|
||||
app.kubernetes.io/name: aerodrome-prometheus
|
||||
cluster: aerodrome
|
||||
annotations:
|
||||
notifications.argoproj.io/subscribe.on-sync-failed.gotify: ""
|
||||
notifications.argoproj.io/subscribe.on-health-degraded.gotify: ""
|
||||
spec:
|
||||
project: homelab
|
||||
source:
|
||||
repoURL: "https://git.snarfnet.net/Snarf/homelab-gitops.git"
|
||||
targetRevision: HEAD
|
||||
path: manifests/aerodrome/prometheus
|
||||
destination:
|
||||
server: "https://140.44.4.11:6443"
|
||||
namespace: default
|
||||
syncPolicy:
|
||||
automated:
|
||||
prune: false
|
||||
selfHeal: true
|
||||
syncOptions:
|
||||
- CreateNamespace=true
|
||||
- ServerSideApply=true
|
||||
523
manifests/aerodrome/grafana/ai-stack-dashboard.json
Normal file
523
manifests/aerodrome/grafana/ai-stack-dashboard.json
Normal file
@@ -0,0 +1,523 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": "-- Grafana --",
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": null,
|
||||
"links": [],
|
||||
"liveNow": false,
|
||||
"panels": [
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 100,
|
||||
"title": "\u26a1 AI Stack Executive Overview & Cloud Cost Offsets",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 1,
|
||||
"title": "Prompt Tokens Processed (In)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 0,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(vllm:prompt_tokens_total)",
|
||||
"legendFormat": "Prompt Tokens",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "blue",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "locale"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"title": "Generation Tokens Synthesized (Out)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 6,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(vllm:generation_tokens_total)",
|
||||
"legendFormat": "Gen Tokens",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "locale"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 3,
|
||||
"title": "Homelab Cloud API Savings ($USD)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 12,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "(sum(vllm:prompt_tokens_total) * 0.00000025) + (sum(vllm:generation_tokens_total) * 0.00000100)",
|
||||
"legendFormat": "Net Savings",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "currencyUSD"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "background",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 4,
|
||||
"title": "Active GPU Cluster Health",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 18,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "count(DCGM_FI_DEV_GPU_TEMP)",
|
||||
"legendFormat": "Tesla T4s Active",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "purple",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 5
|
||||
},
|
||||
"id": 200,
|
||||
"title": "\ud83c\udfae Tesla T4 GPU Cluster Telemetry (Harvester HCI Spacedock)",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 5,
|
||||
"title": "GPU Core Utilization (%)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 6
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_GPU_UTIL",
|
||||
"legendFormat": "{{Hostname}} ({{modelName}})",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 20
|
||||
},
|
||||
"unit": "percent",
|
||||
"min": 0,
|
||||
"max": 100
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 6,
|
||||
"title": "VRAM Memory Allocated & Used",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 6
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_FB_USED * 1024 * 1024",
|
||||
"legendFormat": "{{Hostname}} VRAM Used",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "(DCGM_FI_DEV_FB_USED + DCGM_FI_DEV_FB_FREE) * 1024 * 1024",
|
||||
"legendFormat": "{{Hostname}} VRAM Total",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "bytes"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 7,
|
||||
"title": "GPU Thermal Profile (\u00b0C)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 14
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_GPU_TEMP",
|
||||
"legendFormat": "{{Hostname}} (\u00b0C)",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 10
|
||||
},
|
||||
"unit": "celsius",
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "yellow",
|
||||
"value": 65
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 8,
|
||||
"title": "GPU Power Draw (Watts)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 14
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_POWER_USAGE",
|
||||
"legendFormat": "{{Hostname}} (Watts)",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "watt"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 21
|
||||
},
|
||||
"id": 300,
|
||||
"title": "\ud83e\udde0 vLLM Real-Time Inference Dynamics & Token Velocity",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 9,
|
||||
"title": "Token Generation Velocity (Tokens / Sec)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 22
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(rate(vllm:generation_tokens_total[2m])) by (model_name)",
|
||||
"legendFormat": "{{model_name}} (Out)",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "sum(rate(vllm:prompt_tokens_total[2m])) by (model_name)",
|
||||
"legendFormat": "{{model_name}} (In)",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 25
|
||||
},
|
||||
"unit": "cps"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 10,
|
||||
"title": "Inference Concurrency & Request Backlog",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 22
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "vllm:num_requests_running",
|
||||
"legendFormat": "{{model_name}} Running",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "vllm:num_requests_waiting",
|
||||
"legendFormat": "{{model_name}} Queued",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "stepAfter",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 11,
|
||||
"title": "GPU KV Cache Memory Saturation",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 30
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "vllm:gpu_cache_usage_factor * 100",
|
||||
"legendFormat": "{{model_name}} Cache Usage",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 30
|
||||
},
|
||||
"unit": "percent",
|
||||
"min": 0,
|
||||
"max": 100
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 12,
|
||||
"title": "NVIDIA Triton & Qdrant Operations",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 30
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "rate(nv_inference_request_success[2m])",
|
||||
"legendFormat": "Triton Success/s",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "cpu_cores_used",
|
||||
"legendFormat": "Qdrant CPU Cores",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth"
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
}
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"ai",
|
||||
"vllm",
|
||||
"dcgm",
|
||||
"gpu",
|
||||
"snarfnet"
|
||||
],
|
||||
"templating": {
|
||||
"list": []
|
||||
},
|
||||
"time": {
|
||||
"from": "now-6h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m"
|
||||
]
|
||||
},
|
||||
"timezone": "browser",
|
||||
"title": "SnarfNet AI Stack & Inference Intelligence",
|
||||
"uid": "snarfnet-ai-stack",
|
||||
"version": 1
|
||||
}
|
||||
@@ -111,6 +111,15 @@ spec:
|
||||
subPath: grafana.ini
|
||||
- mountPath: /var/lib/grafana
|
||||
name: storage
|
||||
- mountPath: /etc/grafana/provisioning/datasources/datasources.yaml
|
||||
name: datasources-config
|
||||
subPath: datasources.yaml
|
||||
- mountPath: /etc/grafana/provisioning/dashboards/dashboards.yaml
|
||||
name: dashboards-provider-config
|
||||
subPath: dashboards.yaml
|
||||
- mountPath: /var/lib/grafana-dashboards/ai-stack-dashboard.json
|
||||
name: ai-stack-dashboard
|
||||
subPath: ai-stack-dashboard.json
|
||||
dnsPolicy: ClusterFirst
|
||||
enableServiceLinks: true
|
||||
initContainers:
|
||||
@@ -154,6 +163,18 @@ spec:
|
||||
- name: storage
|
||||
persistentVolumeClaim:
|
||||
claimName: grafana
|
||||
- configMap:
|
||||
defaultMode: 420
|
||||
name: grafana-datasources
|
||||
name: datasources-config
|
||||
- configMap:
|
||||
defaultMode: 420
|
||||
name: grafana-dashboards-provider
|
||||
name: dashboards-provider-config
|
||||
- configMap:
|
||||
defaultMode: 420
|
||||
name: grafana-dashboard-ai-stack
|
||||
name: ai-stack-dashboard
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: Service
|
||||
@@ -250,3 +271,575 @@ spec:
|
||||
- hosts:
|
||||
- grafana.snarfnet.net
|
||||
secretName: firefly-tls-cert
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: grafana-datasources
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: grafana
|
||||
data:
|
||||
datasources.yaml: |
|
||||
apiVersion: 1
|
||||
datasources:
|
||||
- name: Prometheus
|
||||
type: prometheus
|
||||
access: proxy
|
||||
url: http://prometheus:9090
|
||||
isDefault: true
|
||||
jsonData:
|
||||
timeInterval: 15s
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: grafana-dashboards-provider
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: grafana
|
||||
data:
|
||||
dashboards.yaml: |
|
||||
apiVersion: 1
|
||||
providers:
|
||||
- name: 'ai-stack-provider'
|
||||
orgId: 1
|
||||
folder: 'AI Stack & Inference'
|
||||
type: file
|
||||
disableDeletion: false
|
||||
editable: true
|
||||
options:
|
||||
path: /var/lib/grafana-dashboards
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: grafana-dashboard-ai-stack
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: grafana
|
||||
data:
|
||||
ai-stack-dashboard.json: |
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": "-- Grafana --",
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": null,
|
||||
"links": [],
|
||||
"liveNow": false,
|
||||
"panels": [
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 100,
|
||||
"title": "\u26a1 AI Stack Executive Overview & Cloud Cost Offsets",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 1,
|
||||
"title": "Prompt Tokens Processed (In)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 0,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(vllm:prompt_tokens_total)",
|
||||
"legendFormat": "Prompt Tokens",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "blue",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "locale"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"title": "Generation Tokens Synthesized (Out)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 6,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(vllm:generation_tokens_total)",
|
||||
"legendFormat": "Gen Tokens",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "locale"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 3,
|
||||
"title": "Homelab Cloud API Savings ($USD)",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 12,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "(sum(vllm:prompt_tokens_total) * 0.00000025) + (sum(vllm:generation_tokens_total) * 0.00000100)",
|
||||
"legendFormat": "Net Savings",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "currencyUSD"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "background",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 4,
|
||||
"title": "Active GPU Cluster Health",
|
||||
"type": "stat",
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 6,
|
||||
"x": 18,
|
||||
"y": 1
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "count(DCGM_FI_DEV_GPU_TEMP)",
|
||||
"legendFormat": "Tesla T4s Active",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "thresholds"
|
||||
},
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "purple",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
},
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "center"
|
||||
}
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 5
|
||||
},
|
||||
"id": 200,
|
||||
"title": "\ud83c\udfae Tesla T4 GPU Cluster Telemetry (Harvester HCI Spacedock)",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 5,
|
||||
"title": "GPU Core Utilization (%)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 6
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_GPU_UTIL",
|
||||
"legendFormat": "{{Hostname}} ({{modelName}})",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 20
|
||||
},
|
||||
"unit": "percent",
|
||||
"min": 0,
|
||||
"max": 100
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 6,
|
||||
"title": "VRAM Memory Allocated & Used",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 6
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_FB_USED * 1024 * 1024",
|
||||
"legendFormat": "{{Hostname}} VRAM Used",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "(DCGM_FI_DEV_FB_USED + DCGM_FI_DEV_FB_FREE) * 1024 * 1024",
|
||||
"legendFormat": "{{Hostname}} VRAM Total",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "bytes"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 7,
|
||||
"title": "GPU Thermal Profile (\u00b0C)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 14
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_GPU_TEMP",
|
||||
"legendFormat": "{{Hostname}} (\u00b0C)",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 10
|
||||
},
|
||||
"unit": "celsius",
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "yellow",
|
||||
"value": 65
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 8,
|
||||
"title": "GPU Power Draw (Watts)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 14
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "DCGM_FI_DEV_POWER_USAGE",
|
||||
"legendFormat": "{{Hostname}} (Watts)",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "watt"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"collapsed": false,
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 21
|
||||
},
|
||||
"id": 300,
|
||||
"title": "\ud83e\udde0 vLLM Real-Time Inference Dynamics & Token Velocity",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"id": 9,
|
||||
"title": "Token Generation Velocity (Tokens / Sec)",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 22
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(rate(vllm:generation_tokens_total[2m])) by (model_name)",
|
||||
"legendFormat": "{{model_name}} (Out)",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "sum(rate(vllm:prompt_tokens_total[2m])) by (model_name)",
|
||||
"legendFormat": "{{model_name}} (In)",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 25
|
||||
},
|
||||
"unit": "cps"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 10,
|
||||
"title": "Inference Concurrency & Request Backlog",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 22
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "vllm:num_requests_running",
|
||||
"legendFormat": "{{model_name}} Running",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "vllm:num_requests_waiting",
|
||||
"legendFormat": "{{model_name}} Queued",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "stepAfter",
|
||||
"fillOpacity": 15
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 11,
|
||||
"title": "GPU KV Cache Memory Saturation",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 30
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "vllm:gpu_cache_usage_factor * 100",
|
||||
"legendFormat": "{{model_name}} Cache Usage",
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth",
|
||||
"fillOpacity": 30
|
||||
},
|
||||
"unit": "percent",
|
||||
"min": 0,
|
||||
"max": 100
|
||||
}
|
||||
}
|
||||
},
|
||||
{
|
||||
"id": 12,
|
||||
"title": "NVIDIA Triton & Qdrant Operations",
|
||||
"type": "timeseries",
|
||||
"gridPos": {
|
||||
"h": 7,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 30
|
||||
},
|
||||
"datasource": "Prometheus",
|
||||
"targets": [
|
||||
{
|
||||
"expr": "rate(nv_inference_request_success[2m])",
|
||||
"legendFormat": "Triton Success/s",
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"expr": "cpu_cores_used",
|
||||
"legendFormat": "Qdrant CPU Cores",
|
||||
"refId": "B"
|
||||
}
|
||||
],
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"drawStyle": "line",
|
||||
"lineInterpolation": "smooth"
|
||||
},
|
||||
"unit": "short"
|
||||
}
|
||||
}
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"ai",
|
||||
"vllm",
|
||||
"dcgm",
|
||||
"gpu",
|
||||
"snarfnet"
|
||||
],
|
||||
"templating": {
|
||||
"list": []
|
||||
},
|
||||
"time": {
|
||||
"from": "now-6h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m"
|
||||
]
|
||||
},
|
||||
"timezone": "browser",
|
||||
"title": "SnarfNet AI Stack & Inference Intelligence",
|
||||
"uid": "snarfnet-ai-stack",
|
||||
"version": 1
|
||||
}
|
||||
|
||||
182
manifests/aerodrome/prometheus/prometheus.yaml
Normal file
182
manifests/aerodrome/prometheus/prometheus.yaml
Normal file
@@ -0,0 +1,182 @@
|
||||
apiVersion: v1
|
||||
kind: PersistentVolumeClaim
|
||||
metadata:
|
||||
name: prometheus-storage
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
spec:
|
||||
accessModes:
|
||||
- ReadWriteOnce
|
||||
storageClassName: longhorn
|
||||
resources:
|
||||
requests:
|
||||
storage: 50Gi
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: prometheus-config
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
data:
|
||||
prometheus.yml: |
|
||||
global:
|
||||
scrape_interval: 15s
|
||||
scrape_timeout: 10s
|
||||
evaluation_interval: 15s
|
||||
|
||||
scrape_configs:
|
||||
- job_name: 'prometheus'
|
||||
static_configs:
|
||||
- targets: ['localhost:9090']
|
||||
|
||||
- job_name: 'vllm-qwen'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.84:30800']
|
||||
labels:
|
||||
model: 'Qwen2.5-7B-Instruct'
|
||||
node: 'snaisvr44hrvstr04'
|
||||
role: 'general-instruction'
|
||||
|
||||
- job_name: 'vllm-gemma4'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.82:30804']
|
||||
labels:
|
||||
model: 'Gemma-4-12B-it'
|
||||
node: 'snaisvr44hrvstr02'
|
||||
role: 'coding-agent'
|
||||
|
||||
- job_name: 'vllm-deepseek-r1'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.81:30805']
|
||||
labels:
|
||||
model: 'DeepSeek-R1-Distill-Qwen-14B'
|
||||
node: 'snaisvr44hrvstr01'
|
||||
role: 'reasoning-planner'
|
||||
|
||||
- job_name: 'triton-inference-server'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.83:30803']
|
||||
labels:
|
||||
app: 'triton-inference-server'
|
||||
node: 'snaisvr44hrvstr03'
|
||||
model: 'decision_engine'
|
||||
|
||||
- job_name: 'qdrant'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.82:30633']
|
||||
labels:
|
||||
app: 'qdrant'
|
||||
node: 'snaisvr44hrvstr02'
|
||||
|
||||
- job_name: 'nvidia-dcgm'
|
||||
metrics_path: /metrics
|
||||
static_configs:
|
||||
- targets: ['140.44.4.81:9400']
|
||||
labels:
|
||||
node: 'snaisvr44hrvstr01'
|
||||
gpu_model: 'Tesla T4'
|
||||
- targets: ['140.44.4.82:9400']
|
||||
labels:
|
||||
node: 'snaisvr44hrvstr02'
|
||||
gpu_model: 'Tesla T4'
|
||||
- targets: ['140.44.4.83:9400']
|
||||
labels:
|
||||
node: 'snaisvr44hrvstr03'
|
||||
gpu_model: 'Tesla T4'
|
||||
- targets: ['140.44.4.84:9400']
|
||||
labels:
|
||||
node: 'snaisvr44hrvstr04'
|
||||
gpu_model: 'Tesla T4'
|
||||
---
|
||||
apiVersion: apps/v1
|
||||
kind: Deployment
|
||||
metadata:
|
||||
name: prometheus
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
spec:
|
||||
replicas: 1
|
||||
strategy:
|
||||
type: Recreate
|
||||
selector:
|
||||
matchLabels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
template:
|
||||
metadata:
|
||||
labels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
spec:
|
||||
securityContext:
|
||||
fsGroup: 65534
|
||||
runAsUser: 65534
|
||||
runAsNonRoot: true
|
||||
containers:
|
||||
- name: prometheus
|
||||
image: prom/prometheus:v2.54.1
|
||||
imagePullPolicy: IfNotPresent
|
||||
args:
|
||||
- "--config.file=/etc/prometheus/prometheus.yml"
|
||||
- "--storage.tsdb.path=/prometheus"
|
||||
- "--storage.tsdb.retention.time=90d"
|
||||
- "--web.enable-lifecycle"
|
||||
- "--web.console.libraries=/usr/share/prometheus/console_libraries"
|
||||
- "--web.console.templates=/usr/share/prometheus/consoles"
|
||||
ports:
|
||||
- name: web
|
||||
containerPort: 9090
|
||||
resources:
|
||||
requests:
|
||||
cpu: "500m"
|
||||
memory: "1Gi"
|
||||
limits:
|
||||
cpu: "2"
|
||||
memory: "4Gi"
|
||||
readinessProbe:
|
||||
httpGet:
|
||||
path: /-/ready
|
||||
port: 9090
|
||||
initialDelaySeconds: 10
|
||||
periodSeconds: 5
|
||||
livenessProbe:
|
||||
httpGet:
|
||||
path: /-/healthy
|
||||
port: 9090
|
||||
initialDelaySeconds: 30
|
||||
periodSeconds: 15
|
||||
volumeMounts:
|
||||
- name: config
|
||||
mountPath: /etc/prometheus
|
||||
- name: storage
|
||||
mountPath: /prometheus
|
||||
volumes:
|
||||
- name: config
|
||||
configMap:
|
||||
name: prometheus-config
|
||||
- name: storage
|
||||
persistentVolumeClaim:
|
||||
claimName: prometheus-storage
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: Service
|
||||
metadata:
|
||||
name: prometheus
|
||||
namespace: default
|
||||
labels:
|
||||
app.kubernetes.io/name: prometheus
|
||||
spec:
|
||||
type: ClusterIP
|
||||
selector:
|
||||
app.kubernetes.io/name: prometheus
|
||||
ports:
|
||||
- name: web
|
||||
port: 9090
|
||||
targetPort: 9090
|
||||
@@ -609,3 +609,53 @@ spec:
|
||||
port: 8002
|
||||
targetPort: 8002
|
||||
nodePort: 30803
|
||||
|
||||
---
|
||||
# ==============================================================================
|
||||
# 4. NVIDIA DCGM EXPORTER (Tesla T4 Telemetry: Utilization, VRAM, Power, Temp)
|
||||
# ==============================================================================
|
||||
apiVersion: apps/v1
|
||||
kind: DaemonSet
|
||||
metadata:
|
||||
name: dcgm-exporter
|
||||
namespace: ai-stack
|
||||
labels:
|
||||
app.kubernetes.io/name: dcgm-exporter
|
||||
spec:
|
||||
selector:
|
||||
matchLabels:
|
||||
app.kubernetes.io/name: dcgm-exporter
|
||||
template:
|
||||
metadata:
|
||||
labels:
|
||||
app.kubernetes.io/name: dcgm-exporter
|
||||
spec:
|
||||
hostNetwork: true
|
||||
containers:
|
||||
- name: dcgm-exporter
|
||||
image: nvcr.io/nvidia/k8s/dcgm-exporter:3.3.8-3.6.0-ubuntu22.04
|
||||
env:
|
||||
- name: DCGM_EXPORTER_LISTEN
|
||||
value: ":9400"
|
||||
- name: DCGM_EXPORTER_KUBERNETES
|
||||
value: "false"
|
||||
- name: LD_LIBRARY_PATH
|
||||
value: "/usr/local/nvidia/lib64"
|
||||
securityContext:
|
||||
privileged: true
|
||||
ports:
|
||||
- name: metrics
|
||||
containerPort: 9400
|
||||
volumeMounts:
|
||||
- name: dev
|
||||
mountPath: /dev
|
||||
- name: nvidia-lib
|
||||
mountPath: /usr/local/nvidia/lib64
|
||||
readOnly: true
|
||||
volumes:
|
||||
- name: dev
|
||||
hostPath:
|
||||
path: /dev
|
||||
- name: nvidia-lib
|
||||
hostPath:
|
||||
path: /usr/local/nvidia/lib64
|
||||
|
||||
Reference in New Issue
Block a user