[2026-06-01] Fix duplicate bot startup guards

scripts/_watchdog.py

@@ -1,44 +1,83 @@
-"""
-watchdog.py — 봇 생존 감시 + 자동 재시작
-5분마다 실행 (작업 스케줄러) / 장중(09:00~15:10)에만 작동
-"""
-import asyncio, os, subprocess, sys
+import asyncio
+import os
+import subprocess
+import sys
 from datetime import datetime
+from pathlib import Path
 
-PROJECT = r'C:\Users\whdwo\OneDrive\바탕 화면\stockbot_v3'
-PID_FILE = os.path.join(PROJECT, 'logs', 'bot.pid')
+
+PROJECT = Path(__file__).resolve().parents[1]
+PID_FILE = PROJECT / "logs" / "bot.pid"
 
 os.chdir(PROJECT)
-sys.path.insert(0, '.')
+sys.path.insert(0, str(PROJECT))
+
 from app.main import load_env
+
 load_env()
 
 
-def is_process_alive(pid: int) -> bool:
-    r = subprocess.run(
-        ['tasklist', '/FI', f'PID eq {pid}', '/NH'],
-        capture_output=True, text=True,
+def _is_process_alive(pid: int) -> bool:
+    result = subprocess.run(
+        ["tasklist", "/FI", f"PID eq {pid}", "/NH"],
+        capture_output=True,
+        text=True,
     )
-    return str(pid) in r.stdout
+    return str(pid) in result.stdout
 
 
-def get_pid() -> int | None:
+def _find_bot_pids() -> list[int]:
+    command = (
+        "Get-CimInstance Win32_Process | "
+        "Where-Object { $_.Name -like 'python*' -and "
+        "($_.CommandLine -like '*app/main.py*' "
+        "-or $_.CommandLine -like '*app\\\\main.py*') } | "
+        "Select-Object -ExpandProperty ProcessId"
+    )
+    result = subprocess.run(
+        ["powershell", "-NoProfile", "-Command", command],
+        capture_output=True,
+        text=True,
+    )
+    pids: list[int] = []
+    for line in result.stdout.splitlines():
+        line = line.strip()
+        if line.isdigit():
+            pids.append(int(line))
+    return pids
+
+
+def _get_pid() -> int | None:
     try:
-        return int(open(PID_FILE).read().strip())
+        return int(PID_FILE.read_text(encoding="utf-8").strip())
     except Exception:
         return None
 
 
-def restart_bot() -> int:
-    proc = subprocess.Popen(
-        [sys.executable, 'app/main.py'],
-        creationflags=subprocess.DETACHED_PROCESS | subprocess.CREATE_NEW_PROCESS_GROUP,
-        stdout=open('logs/bot_stderr.log', 'a', encoding='utf-8'),
-        stderr=subprocess.STDOUT,
-        close_fds=True,
-    )
-    with open(PID_FILE, 'w') as f:
-        f.write(str(proc.pid))
+def _write_pid(pid: int) -> None:
+    PID_FILE.parent.mkdir(exist_ok=True)
+    PID_FILE.write_text(str(pid), encoding="utf-8")
+
+
+def _restart_bot() -> int:
+    env = os.environ.copy()
+    env["PYTHONUNBUFFERED"] = "1"
+    creationflags = 0
+    if hasattr(subprocess, "DETACHED_PROCESS"):
+        creationflags = subprocess.DETACHED_PROCESS | subprocess.CREATE_NEW_PROCESS_GROUP
+
+    log_path = PROJECT / "logs" / "bot_stderr.log"
+    with open(log_path, "a", encoding="utf-8") as log:
+        proc = subprocess.Popen(
+            [sys.executable, "-u", "app/main.py"],
+            cwd=PROJECT,
+            creationflags=creationflags,
+            stdout=log,
+            stderr=subprocess.STDOUT,
+            close_fds=True,
+            env=env,
+        )
+    _write_pid(proc.pid)
     return proc.pid
 
 
@@ -46,35 +85,33 @@ async def main():
     now = datetime.now()
     now_str = now.strftime("%H:%M")
 
-    # 장 외 시간은 체크 안 함
     if not ("09:00" <= now_str <= "15:10"):
-        print(f"[{now_str}] 장 외 시간 — 워치독 종료")
+        print(f"[{now_str}] outside trading window - watchdog skipped")
         return
 
     from app.monitor.notifier import send
 
-    pid = get_pid()
+    pid = _get_pid()
+    if pid is not None and _is_process_alive(pid):
+        print(f"[{now_str}] bot running PID={pid}")
+        return
 
-    if pid is None:
-        msg = f"[경고] 봇 PID 파일 없음 — 봇이 실행되지 않은 상태입니다 ({now_str})"
+    live_pids = _find_bot_pids()
+    if live_pids:
+        recovered_pid = live_pids[0]
+        _write_pid(recovered_pid)
+        msg = f"[복구] bot.pid corrected to running bot PID={recovered_pid} ({now_str})"
         print(msg)
         await send(msg)
-        new_pid = restart_bot()
-        await send(f"[복구] 봇 자동 재시작 완료 PID={new_pid}")
         return
 
-    if is_process_alive(pid):
-        print(f"[{now_str}] 봇 정상 실행 중 PID={pid}")
-        return
-
-    # 봇이 죽어있음
-    msg = f"[긴급] 봇 프로세스 종료 감지 (PID={pid}) — 자동 재시작 시도"
+    msg = f"[긴급] bot process not found (pid={pid}) - restarting"
     print(msg)
     await send(msg)
 
-    new_pid = restart_bot()
-    await send(f"[복구] 봇 자동 재시작 완료 PID={new_pid} ({now_str})")
-    print(f"봇 재시작 완료 PID={new_pid}")
+    new_pid = _restart_bot()
+    await send(f"[복구] bot restarted PID={new_pid} ({now_str})")
+    print(f"bot restarted PID={new_pid}")
 
 
 if __name__ == "__main__":