skill-agent-log-s3/scripts/collect_metrics.py

#!/usr/bin/env python3
"""采集本机运行指标 → JSON (跨平台: Linux/macOS/Windows-WSL)"""
import json, os, subprocess, datetime, socket, sys

WORK = os.path.expanduser("~/.fmode-harness-agent")

def sh(cmd, timeout=15):
    try:
        return subprocess.run(cmd, shell=True, capture_output=True, text=True, timeout=timeout).stdout.strip()
    except Exception as e:
        return f"ERR:{e}"

def collect():
    now = datetime.datetime.now().strftime("%Y-%m-%dT%H:%M:%S+08:00")
    data = {
        "timestamp": now,
        "agent": os.environ.get("AGENT_NAME", socket.gethostname()),
        "host": {
            "platform": sys.platform,
            "hostname": socket.gethostname(),
        },
        "resources": {},
        "concurrency": {},
        "services": {},
        "events": {}
    }

    # --- CPU/内存 (跨平台: /proc/meminfo + uptime) ---
    try:
        meminfo = {}
        for line in open("/proc/meminfo"):
            k, v = line.split(":")
            meminfo[k.strip()] = int(v.strip().split()[0])  # kB
        total = meminfo["MemTotal"]
        avail = meminfo["MemAvailable"]
        data["resources"]["memory"] = {
            "total_mb": round(total / 1024),
            "used_mb": round((total - avail) / 1024),
            "percent": round((total - avail) / total * 100, 1)
        }
        load = open("/proc/loadavg").read().split()
        data["resources"]["cpu_load"] = {"m1": float(load[0]), "m5": float(load[1]), "m15": float(load[2])}
        # CPU%
        cpu = sh("top -bn1 | grep 'Cpu(s)' | awk '{print $2+$4}'")
        data["resources"]["cpu_percent"] = round(float(cpu), 1) if cpu and not cpu.startswith("ERR") else None
    except Exception as e:
        data["resources"]["error"] = str(e)[:100]

    # --- 磁盘 (工作区所在分区) ---
    disk = sh(f"df -B1 {WORK} | tail -1 | awk '{{print $2, $3, $5}}'")
    if disk and not disk.startswith("ERR"):
        t, u, p = disk.split()
        data["resources"]["disk"] = {"total_gb": round(int(t)/1e9,1), "used_gb": round(int(u)/1e9,1), "percent": p}

    # --- 网络 (累计上下行, 取主网卡) ---
    net = sh("cat /proc/net/dev | awk 'NR>2 && $1!~/lo:/ {gsub(\":\",\"\",$1); print $1, $2, $10}' | sort -k2 -rn | head -1")
    if net and not net.startswith("ERR"):
        iface, rx, tx = net.split()
        data["resources"]["network"] = {"iface": iface, "rx_total_mb": round(int(rx)/1e6,1), "tx_total_mb": round(int(tx)/1e6,1)}

    # --- GPU (有则记) ---
    gpu = sh("nvidia-smi --query-gpu=utilization.gpu,memory.used,memory.total --format=csv,noheader,nounits 2>/dev/null")
    if gpu:
        u, mu, mt = gpu.split(", ")
        data["resources"]["gpu"] = {"percent": int(u), "mem_used_mb": int(mu), "mem_total_mb": int(mt)}

    # --- 并发: profile / session / subagent ---
    profiles = sh("ls /opt/data/profiles/ 2>/dev/null | wc -l")
    data["concurrency"]["profiles_total"] = int(profiles) if profiles.isdigit() else None
    # 活跃session(hermes state.db最近30分钟)
    try:
        import sqlite3
        conn = sqlite3.connect("file:/opt/data/state.db?mode=ro", uri=True, timeout=5)
        cur = conn.cursor()
        cur.execute("SELECT COUNT(DISTINCT session_id) FROM messages WHERE timestamp > ?", (time.time()-1800,))
        data["concurrency"]["active_sessions_30min"] = cur.fetchone()[0]
        cur.execute("SELECT COUNT(*) FROM async_delegations WHERE status='running'")
        data["concurrency"]["running_subagents"] = cur.fetchone()[0]
        conn.close()
    except Exception:
        data["concurrency"]["active_sessions_30min"] = None

    # --- 服务存活 ---
    for svc, pat in [("gateway", "hermes gateway"), ("dashboard", "hermes-dashboard"), ("studio", "fmode-studio")]:
        out = sh(f"pgrep -f '{pat}' | wc -l")
        data["services"][svc] = "up" if out.isdigit() and int(out) > 0 else "down"

    # --- 事件: 最近1小时错误日志计数 ---
    errc = sh("find /opt/data/logs -name '*.log' -mmin -60 -exec grep -ci 'error\\|fatal' {} + 2>/dev/null | awk -F: '{s+=$2} END {print s}'")
    data["events"]["errors_last_1h"] = int(errc) if errc.isdigit() else None
    rest = sh("grep -c 'Reconnected' /opt/data/logs/gateway.log 2>/dev/null")
    data["events"]["wecom_reconnects_total"] = int(rest) if rest.isdigit() else None

    return data

if __name__ == "__main__":
    import time
    out = os.path.join(WORK, "tmp")
    os.makedirs(out, exist_ok=True)
    d = collect()
    path = os.path.join(out, "agent-log.json")
    json.dump(d, open(path, "w"), ensure_ascii=False, indent=2)
    print(json.dumps(d, ensure_ascii=False)[:300])
    print(f"\nsaved: {path}")