#!/usr/bin/env python3
"""fix_gateway.py — 网关健康检查+自愈脚本。

检查 gateway_zhiwei (position-analyst profile) 是否存活，端口8643是否开放。
如果挂了则重启。

执行器调用方式: python3 fix_gateway.py <component_name>
"""
import sys, os, json, subprocess, socket

GATEWAY_PORT = 8643
GATEWAY_PROFILE = "position-analyst"
HERMES_BIN = "/home/hmo/hermes-agent/.venv/bin/python"
GATEWAY_WD = "/home/hmo/hermes-agent"

def tcp_check(host="127.0.0.1", port=8643, timeout=2):
    """TCP端口检查"""
    try:
        s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        s.settimeout(timeout)
        result = s.connect_ex((host, port))
        s.close()
        return result == 0
    except Exception:
        return False

def check_gateway_process():
    """检查对应profile的网关进程"""
    import subprocess
    r = subprocess.run(
        ["pgrep", "-f", f"position-analyst.*gateway.*--replace"],
        capture_output=True, text=True, timeout=5
    )
    pids = [p.strip() for p in r.stdout.strip().split("\n") if p.strip()]
    return pids

def restart_gateway():
    """重启网关"""
    pids = check_gateway_process()
    if pids:
        for pid in pids:
            try:
                os.kill(int(pid), 15)  # SIGTERM
            except:
                pass
        import time
        time.sleep(2)
    
    # 启动新网关（后台）
    subprocess.Popen(
        [HERMES_BIN, "-m", "hermes_cli.main", "-p", GATEWAY_PROFILE, 
         "gateway", "run", "--replace"],
        cwd=GATEWAY_WD,
        stdout=subprocess.DEVNULL,
        stderr=subprocess.DEVNULL,
        start_new_session=True
    )
    return True

def main():
    component = sys.argv[1] if len(sys.argv) > 1 else "gateway_zhiwei"
    
    # 1. TCP端口检查
    port_ok = tcp_check(port=GATEWAY_PORT)
    
    # 2. 进程检查
    pids = check_gateway_process()
    proc_ok = len(pids) > 0
    
    if port_ok and proc_ok:
        print(f"[OK] {component}: TCP端口{GATEWAY_PORT}开放, 进程={','.join(pids)}")
        return 0
    
    # 异常情况
    print(f"[DOWN] {component}: TCP端口={port_ok}, 进程={proc_ok} pids={pids}")
    print(f"[FIX] 正在重启网关...")
    
    if restart_gateway():
        import time
        time.sleep(5)
        port_ok = tcp_check(port=GATEWAY_PORT)
        pids = check_gateway_process()
        if port_ok and pids:
            print(f"[OK] {component}: 重启成功 端口={port_ok} pids={pids}")
            return 0
        else:
            print(f"[FAIL] {component}: 重启后仍未恢复 port={port_ok} pids={pids}")
            return 1
    
    return 1

if __name__ == "__main__":
    sys.exit(main())