""" 远程主机系统监控模块 通过 SSH 一次性采集 CPU/内存/磁盘/网络/负载指标。 Linux 用 /proc 和常用命令;macOS/BSD 走兼容路径。 """ import re import time from typing import Optional from .ssh_client import SSHConnection class SystemMonitor: """远程主机的资源监控器(数据全部从 SSH 通道采集,不依赖 agent)""" # 一次性获取所有指标的脚本(Linux) _LINUX_METRICS_SCRIPT = r""" echo "===CPU===" # 第一次采样 1 秒间隔,用来计算差值 read cpu_user cpu_nice cpu_system cpu_idle cpu_iowait cpu_irq cpu_softirq cpu_steal < /proc/stat sleep 1 read cpu_user2 cpu_nice2 cpu_system2 cpu_idle2 cpu_iowait2 cpu_irq2 cpu_softirq2 cpu_steal2 < /proc/stat total1=$((cpu_user+cpu_nice+cpu_system+cpu_idle+cpu_iowait+cpu_irq+cpu_softirq+cpu_steal)) total2=$((cpu_user2+cpu_nice2+cpu_system2+cpu_idle2+cpu_iowait2+cpu_irq2+cpu_softirq2+cpu_steal2)) idle1=$cpu_idle; idle2=$cpu_idle2 dt=$((total2-total1)); di=$((idle2-idle1)) if [ $dt -gt 0 ]; then usage=$(( (1000*(dt-di)/dt+5)/10 )); else usage=0; fi echo "CPU_USAGE=$usage" echo "CPU_CORES=$(nproc 2>/dev/null || echo 1)" echo "LOAD=$(cat /proc/loadavg | awk '{print $1,$2,$3}')" echo "UPTIME=$(awk '{printf "%.0f",$1}' /proc/uptime)" echo "===MEM===" mem_total=$(awk '/MemTotal/{print $2}' /proc/meminfo) mem_avail=$(awk '/MemAvailable/{print $2}' /proc/meminfo) swap_total=$(awk '/SwapTotal/{print $2}' /proc/meminfo) swap_free=$(awk '/SwapFree/{print $2}' /proc/meminfo) if [ -z "$mem_avail" ]; then mem_avail=$((mem_total - $(awk '/^(Buffers|Cached|SReclaimable):/{s+=$2} END{print s}' /proc/meminfo))); fi used=$((mem_total - mem_avail)) echo "MEM_TOTAL=$mem_total" echo "MEM_USED=$used" echo "MEM_AVAIL=$mem_avail" echo "SWAP_TOTAL=$swap_total" echo "SWAP_USED=$((swap_total-swap_free))" echo "===DISK===" df -PB1 -x tmpfs -x devtmpfs 2>/dev/null | awk 'NR>1 {printf "DISK|%s|%d|%d|%s\n",$NF,$2,$3,$5}' echo "===NET===" for iface in $(ls /sys/class/net/ 2>/dev/null | grep -v lo); do rx=$(cat /sys/class/net/$iface/statistics/rx_bytes 2>/dev/null || echo 0) tx=$(cat /sys/class/net/$iface/statistics/tx_bytes 2>/dev/null || echo 0) echo "NET|$iface|$rx|$tx" done echo "===HOST===" echo "HOSTNAME=$(hostname)" echo "KERNEL=$(uname -r)" echo "OS=$(. /etc/os-release 2>/dev/null && echo "$PRETTY_NAME" || uname -s)" """ @staticmethod def _parse_kv(text: str, key: str, default: str = "0") -> str: """从 KEY=VALUE 行中取值""" m = re.search(rf"^{re.escape(key)}=(.+)$", text, re.MULTILINE) return m.group(1).strip() if m else default @classmethod def collect(cls, conn: SSHConnection) -> dict: """采集一次指标;返回 dict""" empty = { "cpu": 0.0, "cores": 1, "load1": 0, "load5": 0, "load15": 0, "uptime": 0, "hostname": "", "kernel": "", "os": "", "mem_total": 0, "mem_used": 0, "mem_percent": 0.0, "swap_total": 0, "swap_used": 0, "disks": [], "net": [], "ts": time.time(), } if not conn or not conn.connected: return empty code, out, err = conn.exec_command(cls._LINUX_METRICS_SCRIPT, timeout=10) if code != 0 or not out: empty["error"] = err or "采集失败" return empty result = dict(empty) result["hostname"] = cls._parse_kv(out, "HOSTNAME") result["kernel"] = cls._parse_kv(out, "KERNEL") result["os"] = cls._parse_kv(out, "OS") try: result["cpu"] = float(cls._parse_kv(out, "CPU_USAGE")) except ValueError: pass try: result["cores"] = int(cls._parse_kv(out, "CPU_CORES", "1")) except ValueError: pass load = cls._parse_kv(out, "LOAD", "0 0 0").split() try: result["load1"] = float(load[0]) result["load5"] = float(load[1]) if len(load) > 1 else 0 result["load15"] = float(load[2]) if len(load) > 2 else 0 except (ValueError, IndexError): pass try: result["uptime"] = int(cls._parse_kv(out, "UPTIME")) except ValueError: pass try: mt = int(cls._parse_kv(out, "MEM_TOTAL")) mu = int(cls._parse_kv(out, "MEM_USED")) result["mem_total"] = mt result["mem_used"] = mu result["mem_percent"] = (mu / mt * 100) if mt > 0 else 0.0 result["swap_total"] = int(cls._parse_kv(out, "SWAP_TOTAL")) result["swap_used"] = int(cls._parse_kv(out, "SWAP_USED")) except ValueError: pass result["disks"] = [] for line in out.splitlines(): if line.startswith("DISK|"): _, mount, total, used, percent = line.split("|", 4) try: result["disks"].append({ "mount": mount, "total": int(total), "used": int(used), "percent": int(percent.rstrip("%")), }) except ValueError: continue result["net"] = [] for line in out.splitlines(): if line.startswith("NET|"): _, name, rx, tx = line.split("|", 3) try: result["net"].append({ "iface": name, "rx": int(rx), "tx": int(tx), }) except ValueError: continue return result @staticmethod def format_bytes(n: int) -> str: """人类可读字节数""" n = float(n) for unit in ("B", "KB", "MB", "GB", "TB", "PB"): if n < 1024: return f"{n:.1f}{unit}" n /= 1024 return f"{n:.1f}EB" @staticmethod def format_uptime(seconds: int) -> str: seconds = int(seconds) d, rem = divmod(seconds, 86400) h, rem = divmod(rem, 3600) m, s = divmod(rem, 60) if d: return f"{d}天{h}小时" if h: return f"{h}小时{m}分" if m: return f"{m}分{s}秒" return f"{s}秒"