Add GPU utilization monitoring (NVIDIA/Intel) and dropshell service discovery
All checks were successful
Build-Publish / build (linux/amd64) (push) Successful in 16s
Build-Publish / build (linux/arm64) (push) Successful in 38s
Build-Publish / create-manifest (push) Successful in 2s
Build-Publish / publish-template (push) Successful in 16s

This commit is contained in:
j
2026-03-18 22:34:46 +13:00
parent 975127a694
commit db5cbf99e1
3 changed files with 173 additions and 8 deletions

View File

@@ -143,6 +143,45 @@ while read -r line; do
gpu_idx=$((gpu_idx + 1))
done < <(lspci 2>/dev/null | grep -iE 'vga|3d|display')
# GPU utilization (NVIDIA)
if command -v nvidia-smi &>/dev/null; then
nv_idx=0
while IFS=',' read -r name util mem_used mem_total temp; do
echo "[nvidia_gpu:$nv_idx]"
echo "name=$(echo "$name" | xargs)"
echo "utilization_percent=$(echo "$util" | xargs)"
echo "memory_used_mb=$(echo "$mem_used" | xargs)"
echo "memory_total_mb=$(echo "$mem_total" | xargs)"
echo "temperature=$(echo "$temp" | xargs)"
nv_idx=$((nv_idx + 1))
done < <(nvidia-smi --query-gpu=name,utilization.gpu,memory.used,memory.total,temperature.gpu --format=csv,noheader,nounits 2>/dev/null)
fi
# GPU utilization (Intel)
igpu_cmd=$(command -v intel_gpu_top 2>/dev/null)
if [ -n "$igpu_cmd" ]; then
igpu_prefix=""
[ "$(id -u)" -ne 0 ] && igpu_prefix="sudo"
igpu_idx=0
for drm in /sys/class/drm/card*; do
[ -d "$drm" ] || continue
driver=$(readlink "$drm/device/driver" 2>/dev/null | xargs basename 2>/dev/null)
[ "$driver" = "i915" ] || [ "$driver" = "xe" ] || continue
card=$(basename "$drm")
igpu_raw=$(timeout 2 $igpu_prefix intel_gpu_top -J -s 500 -d /dev/dri/"$card" 2>/dev/null)
if [ -n "$igpu_raw" ]; then
echo "[intel_gpu:$igpu_idx]"
busy=$(echo "$igpu_raw" | grep -oP '"busy"\s*:\s*\K[0-9.]+' | sort -rn | head -1)
echo "utilization_percent=${busy:-0}"
freq=$(echo "$igpu_raw" | grep -oP '"actual"\s*:\s*\K[0-9.]+' | head -1)
[ -n "$freq" ] && echo "frequency_mhz=${freq%.*}"
power=$(echo "$igpu_raw" | grep -oP '"GPU"\s*:\s*\K[0-9.]+' | head -1)
[ -n "$power" ] && echo "power_w=$power"
igpu_idx=$((igpu_idx + 1))
fi
done
fi
# Network interfaces
for iface in $(ls /sys/class/net/ 2>/dev/null); do
[ "$iface" = "lo" ] && continue
@@ -311,6 +350,22 @@ if command -v virsh &>/dev/null; then
done
fi
# Dropshell services
ds_idx=0
for svc_dir in /home/dropshell/dropshell/services/*/; do
[ -d "$svc_dir" ] || continue
svc_env="$svc_dir/config/service.env"
[ -f "$svc_env" ] || continue
svc_name=$(basename "$svc_dir")
echo "[dropshell_service:$ds_idx]"
echo "name=$svc_name"
container_name=$(grep '^CONTAINER_NAME=' "$svc_env" 2>/dev/null | head -1 | cut -d= -f2 | tr -d '"')
[ -n "$container_name" ] && echo "container_name=$container_name"
template=$(grep '^TEMPLATE=' "$svc_env" 2>/dev/null | head -1 | cut -d= -f2 | tr -d '"')
[ -n "$template" ] && echo "template=$template"
ds_idx=$((ds_idx + 1))
done
# Docker containers
if command -v docker &>/dev/null; then
_sudo docker ps -a --format '{{.Names}}\t{{.State}}\t{{.Image}}\t{{.Status}}' 2>/dev/null | while IFS=$'\t' read -r name state image status_text; do