wsgi.py

"""
WSGI entry point for Gunicorn production server.
Usage: gunicorn -w 8 -b 0.0.0.0:5000 wsgi:app

Only one worker starts the scheduler (file-lock based).
"""
import os
import fcntl
import time
from app import app, init_db, crawl_update_all, is_match_day, db_health_check

# Initialize database on first import
init_db()

# File-lock based singleton scheduler: only one gunicorn worker wins the race
LOCK_FILE = "/tmp/short_track_scheduler.lock"

scheduler_started = False

try:
    lock_fd = os.open(LOCK_FILE, os.O_CREAT | os.O_WRONLY, 0o644)
    try:
        fcntl.flock(lock_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
        # We got the lock — start scheduler (only one worker will reach here)
        from apscheduler.schedulers.background import BackgroundScheduler
        _last_non_match_day_run = [0]

        def smart_crawl_update():
            """智能爬取：比赛日5分钟一次，非比赛日7天一次"""
            try:
                match_day = is_match_day()
            except Exception:
                match_day = False

            if match_day:
                print("比赛日 - 执行数据更新")
                crawl_update_all()
            else:
                now = time.time()
                SEVEN_DAYS = 7 * 24 * 3600
                if now - _last_non_match_day_run[0] >= SEVEN_DAYS:
                    print("非比赛日 - 7天兜底更新")
                    crawl_update_all()
                    _last_non_match_day_run[0] = now

        scheduler = BackgroundScheduler()
        scheduler.add_job(
            func=smart_crawl_update,
            trigger="interval",
            minutes=5,
            id="auto_update",
            name="智能自动更新（比赛日5分钟，非比赛日7天）",
            replace_existing=True,
            max_instances=1,
            coalesce=True,
            misfire_grace_time=120,
        )
        scheduler.add_job(
            func=db_health_check,
            trigger="interval",
            minutes=5,
            id="db_health_check",
            name="DB完整性检查+自动恢复",
            replace_existing=True,
            max_instances=1,
            coalesce=True,
            misfire_grace_time=120,
        )
        scheduler.start()
        scheduler_started = True
        print("已启动智能定时更新：比赛日5分钟一次，非比赛日7天一次")
        print("已启动 DB 完整性检查：每5分钟一次")

        # 启动时立即执行一次健康检查
        db_health_check()

    except BlockingIOError:
        # Another worker already holds the lock — skip
        print("调度器已在其他 worker 运行，跳过")
finally:
    # Don't close lock_fd — we need to hold it for the process lifetime
    pass

if __name__ == "__main__":
    app.run()