kubesolo-os/test/benchmark/bench-resources.sh

#!/bin/bash
# bench-resources.sh — Measure KubeSolo OS resource usage
#
# Connects to a running KubeSolo OS instance and measures:
#   - Memory usage (total, used, available, per-process)
#   - Disk usage (rootfs, data partition, containerd)
#   - CPU usage under idle and load
#   - Process count
#   - Container count
#   - Network overhead
#
# Usage:
#   test/benchmark/bench-resources.sh [--ssh-port 2222]
#
# Prerequisites: KubeSolo OS running (e.g. via make dev-vm)
set -euo pipefail

SSH_PORT="${SSH_PORT:-2222}"
SSH_HOST="${SSH_HOST:-localhost}"
SSH_USER="${SSH_USER:-root}"
SSH_OPTS="-o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o LogLevel=ERROR"

while [ $# -gt 0 ]; do
    case "$1" in
        --ssh-port) SSH_PORT="$2"; shift 2 ;;
        --ssh-host) SSH_HOST="$2"; shift 2 ;;
        *) echo "Unknown option: $1" >&2; exit 1 ;;
    esac
done

run_ssh() {
    ssh $SSH_OPTS -p "$SSH_PORT" "$SSH_USER@$SSH_HOST" "$@" 2>/dev/null
}

echo "=== KubeSolo OS Resource Benchmark ===" >&2
echo "Connecting to ${SSH_HOST}:${SSH_PORT}..." >&2

# Check connectivity
if ! run_ssh "true" 2>/dev/null; then
    echo "ERROR: Cannot connect via SSH. Is KubeSolo OS running?" >&2
    echo "Start with: make dev-vm" >&2
    exit 1
fi

echo "" >&2

# --- Memory ---
echo "--- Memory Usage ---" >&2
MEM_INFO=$(run_ssh "cat /proc/meminfo")
MEM_TOTAL=$(echo "$MEM_INFO" | sed -n 's/MemTotal: *\([0-9]*\).*/\1/p')
MEM_FREE=$(echo "$MEM_INFO" | sed -n 's/MemFree: *\([0-9]*\).*/\1/p')
MEM_AVAIL=$(echo "$MEM_INFO" | sed -n 's/MemAvailable: *\([0-9]*\).*/\1/p')
MEM_USED=$((MEM_TOTAL - MEM_FREE))

echo "  Total:     $((MEM_TOTAL / 1024)) MB" >&2
echo "  Used:      $((MEM_USED / 1024)) MB" >&2
echo "  Available: $((MEM_AVAIL / 1024)) MB" >&2
echo "  OS overhead: $((MEM_USED / 1024)) MB ($(( (MEM_USED * 100) / MEM_TOTAL ))%)" >&2
echo "" >&2

# Top memory consumers
echo "--- Top Processes (by RSS) ---" >&2
run_ssh "ps -o pid,rss,comm | sort -k2 -rn | head -10" 2>/dev/null | while read -r line; do
    echo "  $line" >&2
done
echo "" >&2

# --- Disk ---
echo "--- Disk Usage ---" >&2
run_ssh "df -h / /mnt/data 2>/dev/null || df -h /" | while read -r line; do
    echo "  $line" >&2
done
echo "" >&2

# Containerd data
CONTAINERD_SIZE=$(run_ssh "du -sh /var/lib/containerd 2>/dev/null | cut -f1" || echo "N/A")
KUBESOLO_SIZE=$(run_ssh "du -sh /var/lib/kubesolo 2>/dev/null | cut -f1" || echo "N/A")
echo "  containerd data: $CONTAINERD_SIZE" >&2
echo "  kubesolo data:   $KUBESOLO_SIZE" >&2
echo "" >&2

# --- Processes ---
echo "--- Process Count ---" >&2
PROC_COUNT=$(run_ssh "ps | wc -l")
echo "  Total processes: $PROC_COUNT" >&2
echo "" >&2

# --- K8s Status ---
echo "--- Kubernetes Status ---" >&2
NODE_STATUS=$(run_ssh "kubesolo kubectl get nodes -o wide --no-headers 2>/dev/null" || echo "N/A")
POD_COUNT=$(run_ssh "kubesolo kubectl get pods -A --no-headers 2>/dev/null | wc -l" || echo "0")
echo "  Node: $NODE_STATUS" >&2
echo "  Pod count: $POD_COUNT" >&2
echo "" >&2

# --- CPU (5-second sample) ---
echo "--- CPU Usage (5s idle sample) ---" >&2
CPU_IDLE=$(run_ssh "
    read cpu user nice system idle rest < /proc/stat
    sleep 5
    read cpu user2 nice2 system2 idle2 rest2 < /proc/stat
    total=\$((user2 + nice2 + system2 + idle2 - user - nice - system - idle))
    idle_diff=\$((idle2 - idle))
    if [ \$total -gt 0 ]; then
        echo \$((idle_diff * 100 / total))
    else
        echo 0
    fi
" 2>/dev/null || echo "N/A")
echo "  CPU idle: ${CPU_IDLE}%" >&2
echo "  CPU used: $((100 - ${CPU_IDLE:-0}))%" >&2
echo "" >&2

# --- OS Version ---
OS_VERSION=$(run_ssh "cat /etc/kubesolo-os-version 2>/dev/null" || echo "unknown")

# --- Output JSON ---
cat << EOF
{
  "benchmark": "kubesolo-os-resources",
  "os_version": "$OS_VERSION",
  "memory": {
    "total_kb": $MEM_TOTAL,
    "used_kb": $MEM_USED,
    "available_kb": ${MEM_AVAIL:-0},
    "total_mb": $((MEM_TOTAL / 1024)),
    "used_mb": $((MEM_USED / 1024)),
    "available_mb": $((${MEM_AVAIL:-0} / 1024)),
    "overhead_percent": $(( (MEM_USED * 100) / MEM_TOTAL ))
  },
  "disk": {
    "containerd_size": "$CONTAINERD_SIZE",
    "kubesolo_size": "$KUBESOLO_SIZE"
  },
  "processes": {
    "total": $PROC_COUNT
  },
  "kubernetes": {
    "pod_count": $POD_COUNT
  },
  "cpu": {
    "idle_percent": ${CPU_IDLE:-0},
    "used_percent": $((100 - ${CPU_IDLE:-0}))
  }
}
EOF