mirror of
https://github.com/Alkatrazz24/Funk-lab.git
synced 2026-07-08 06:24:42 +02:00
92a882aa94
Ansible roles: - node_exporter : déploie node_exporter v1.9.1 sur storage-01 et gpu-01 avec textfile collector + services systemd. ROCm activé sur gpu-01 uniquement via group_vars. Collecte métriques GPU via sysfs (temp, VRAM, utilisation) — pas de dépendance rocm-smi (remplacé dans ROCm 7.x) - alertmanager_webhook : service Python sur storage-01 (:9093/webhook) reçoit alertes AlertManager et les route vers ask-agent monitor → Hermes Playbooks: node_exporter + alertmanager_webhook ajoutés à storage-01 et gpu-01 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
33 lines
1.4 KiB
Django/Jinja
33 lines
1.4 KiB
Django/Jinja
#!/bin/bash
|
|
# Collecte les métriques GPU AMD via sysfs (pas de rocm-smi requis)
|
|
OUTPUT="{{ node_exporter_textfile_dir }}/rocm.prom"
|
|
TMPFILE="${OUTPUT}.tmp"
|
|
|
|
{
|
|
echo "# HELP rocm_gpu_temperature_celsius GPU temperature in Celsius"
|
|
echo "# TYPE rocm_gpu_temperature_celsius gauge"
|
|
echo "# HELP rocm_gpu_utilization_percent GPU utilization percentage"
|
|
echo "# TYPE rocm_gpu_utilization_percent gauge"
|
|
echo "# HELP rocm_vram_used_bytes VRAM used in bytes"
|
|
echo "# TYPE rocm_vram_used_bytes gauge"
|
|
echo "# HELP rocm_vram_total_bytes VRAM total in bytes"
|
|
echo "# TYPE rocm_vram_total_bytes gauge"
|
|
|
|
for card in /sys/class/drm/card[0-9]; do
|
|
[ -d "$card/device" ] || continue
|
|
idx=$(basename "$card" | tr -d 'card')
|
|
labels="gpu=\"${idx}\",model=\"gfx1031\""
|
|
|
|
temp_raw=$(cat "$card/device/hwmon/hwmon"*/temp1_input 2>/dev/null | head -1)
|
|
[ -n "$temp_raw" ] && echo "rocm_gpu_temperature_celsius{${labels}} $(echo "scale=1; ${temp_raw}/1000" | bc)"
|
|
|
|
util=$(cat "$card/device/gpu_busy_percent" 2>/dev/null)
|
|
[ -n "$util" ] && echo "rocm_gpu_utilization_percent{${labels}} ${util}"
|
|
|
|
vram_used=$(cat "$card/device/mem_info_vram_used" 2>/dev/null)
|
|
[ -n "$vram_used" ] && echo "rocm_vram_used_bytes{${labels}} ${vram_used}"
|
|
|
|
vram_total=$(cat "$card/device/mem_info_vram_total" 2>/dev/null)
|
|
[ -n "$vram_total" ] && echo "rocm_vram_total_bytes{${labels}} ${vram_total}"
|
|
done
|
|
} > "$TMPFILE" && mv "$TMPFILE" "$OUTPUT"
|