wangchuanli
23799b4ea5
* LICENSE —— MIT
* THIRD-PARTY-NOTICES —— 依赖清单、再分发合规说明（含随仓库分发的
Apache ECharts 5.6.0 / Apache-2.0）与自查清单
* CONTRIBUTING.md —— 开发环境、五层验证、必须遵守的不变量、提交规范
* SECURITY.md —— 漏洞私有报告渠道、已有措施、已知非目标
* CODE_OF_CONDUCT.md —— 改编自 Contributor Covenant 2.1
* .github/ —— Bug 报告 / 功能建议表单 + PR 模板
* .editorconfig —— 与 .gitattributes 保持一致
* 全部 Python / Shell 源文件加 SPDX-License-Identifier: MIT 头
* README 增加「开源与许可」章节与许可标识
182 行
6.3 KiB
Python
182 行
6.3 KiB
Python
# -*- coding: utf-8 -*-
|
||
# SPDX-License-Identifier: MIT
|
||
# Copyright (c) 2026 Wang Chuanli
|
||
|
||
"""进程内调度器（手写，不依赖 APScheduler）。
|
||
|
||
为什么不引 APScheduler：
|
||
* 需求只是「每天固定几个时刻跑一次」，一个线程 + 20 秒轮询足够
|
||
* 需要「启动补跑」（程序没开的时候错过了时刻，开机后要补上）
|
||
* 需要和 CLI 共享同一把文件锁，避免两处同时采集
|
||
|
||
单实例保证：
|
||
* Flask 的 reloader 会 fork 两个进程 → 只在 WERKZEUG_RUN_MAIN 里启动
|
||
* 多进程部署时用环境变量 WB_DISABLE_SCHEDULER=1 关掉除一个之外的所有实例
|
||
* 真正防重复采集靠 collect._Lock（文件锁），即使两个调度线程同时触发也只会跑一个
|
||
"""
|
||
import logging
|
||
import os
|
||
import threading
|
||
from datetime import datetime, timedelta
|
||
|
||
from . import collect, db
|
||
|
||
log = logging.getLogger("wb.scheduler")
|
||
SLOT_PREFIX = "slot:" # settings 键：slot:09:00 -> 最近执行的日期
|
||
|
||
|
||
def parse_times(raw):
|
||
"""把 '09:00,17:00' 解析成 ['09:00','17:00']；非法片段直接丢弃。
|
||
|
||
对外公开（config.normalize_setting 用它做写入校验），所以不要改名。
|
||
"""
|
||
out = []
|
||
for part in str(raw or "").replace("；", ",").replace("，", ",").split(","):
|
||
p = part.strip()
|
||
if not p:
|
||
continue
|
||
bits = p.split(":")
|
||
try:
|
||
hh = int(bits[0])
|
||
mm = int(bits[1]) if len(bits) > 1 else 0
|
||
except ValueError:
|
||
continue
|
||
if 0 <= hh <= 23 and 0 <= mm <= 59:
|
||
out.append("%02d:%02d" % (hh, mm))
|
||
return sorted(set(out))
|
||
|
||
|
||
# 兼容旧名（早期版本用 _parse_times）
|
||
_parse_times = parse_times
|
||
|
||
|
||
def slots(conn):
|
||
return parse_times(db.get_setting(conn, "schedule_times"))
|
||
|
||
|
||
def last_run_of_slot(conn, slot):
|
||
return db.get_setting(conn, SLOT_PREFIX + slot, "")
|
||
|
||
|
||
def mark_slot(conn, slot, day):
|
||
db.set_setting(conn, SLOT_PREFIX + slot, day)
|
||
|
||
|
||
def next_run_at(conn, now=None):
|
||
"""下一次计划执行时间（仅按配置推算，不含补跑）。"""
|
||
if not db.get_bool(conn, "schedule_enabled", True):
|
||
return None
|
||
now = now or datetime.now()
|
||
best = None
|
||
for s in slots(conn):
|
||
hh, mm = map(int, s.split(":"))
|
||
cand = now.replace(hour=hh, minute=mm, second=0, microsecond=0)
|
||
if cand <= now:
|
||
cand += timedelta(days=1)
|
||
if best is None or cand < best:
|
||
best = cand
|
||
return best
|
||
|
||
|
||
def due_slots(conn, now=None):
|
||
"""返回此刻应当执行的槽位列表（含启动补跑）。"""
|
||
if not db.get_bool(conn, "schedule_enabled", True):
|
||
return []
|
||
now = now or datetime.now()
|
||
today = now.strftime("%Y-%m-%d")
|
||
# 用 get_int 兜底：catch_up_grace_hours 在后台是自由文本框，
|
||
# 历史上填成 "12h" 会让这里 int() 抛 ValueError，把 /tasks 打成 500。
|
||
grace_hours = db.get_int(conn, "catch_up_grace_hours", 12)
|
||
grace = timedelta(hours=max(1, grace_hours))
|
||
catch_up = db.get_bool(conn, "catch_up", True)
|
||
out = []
|
||
for s in slots(conn):
|
||
hh, mm = map(int, s.split(":"))
|
||
when = now.replace(hour=hh, minute=mm, second=0, microsecond=0)
|
||
if when > now:
|
||
continue # 还没到点
|
||
if last_run_of_slot(conn, s) == today:
|
||
continue # 今天这个槽位已跑过
|
||
if when < now - grace and catch_up:
|
||
continue # 错过太久，不补（避免开机狂刷）
|
||
if when < now - timedelta(seconds=90) and not catch_up:
|
||
continue # 未开启补跑，只认刚到的点
|
||
out.append(s)
|
||
return out
|
||
|
||
|
||
class Scheduler:
|
||
def __init__(self, interval=20):
|
||
self.interval = interval
|
||
self._stop = threading.Event()
|
||
self._thread = None
|
||
|
||
# ---- 生命周期 ----
|
||
def start(self):
|
||
if self._thread and self._thread.is_alive():
|
||
return False
|
||
self._stop.clear()
|
||
self._thread = threading.Thread(target=self._loop, name="wb-scheduler", daemon=True)
|
||
self._thread.start()
|
||
log.info("调度器已启动，轮询间隔 %ss", self.interval)
|
||
return True
|
||
|
||
def stop(self):
|
||
self._stop.set()
|
||
if self._thread:
|
||
self._thread.join(timeout=5)
|
||
|
||
@property
|
||
def running(self):
|
||
return bool(self._thread and self._thread.is_alive())
|
||
|
||
# ---- 主循环 ----
|
||
def _loop(self):
|
||
while not self._stop.is_set():
|
||
try:
|
||
self.tick()
|
||
except Exception as e: # 任何异常都不能让线程死掉
|
||
log.exception("调度 tick 出错：%s", e)
|
||
self._stop.wait(self.interval)
|
||
|
||
def tick(self, now=None):
|
||
conn = db.thread_conn()
|
||
now = now or datetime.now()
|
||
today = now.strftime("%Y-%m-%d")
|
||
for slot in due_slots(conn, now):
|
||
scheduled = now.replace(hour=int(slot[:2]), minute=int(slot[3:]),
|
||
second=0, microsecond=0)
|
||
trigger = "startup" if now - scheduled > timedelta(minutes=5) else "schedule"
|
||
log.info("触发采集：槽位 %s（%s）", slot, trigger)
|
||
mark_slot(conn, slot, today) # 先占位，避免采集失败被无限重试打爆云端
|
||
try:
|
||
r = collect.run_sync(trigger=trigger)
|
||
log.info("采集完成：%s", r["message"])
|
||
except collect.Busy as e:
|
||
log.warning("跳过（%s）", e)
|
||
except Exception as e:
|
||
log.error("采集失败：%s", e)
|
||
return True
|
||
|
||
|
||
_scheduler = None
|
||
|
||
|
||
def get_scheduler(interval=20):
|
||
global _scheduler
|
||
if _scheduler is None:
|
||
_scheduler = Scheduler(interval=interval)
|
||
return _scheduler
|
||
|
||
|
||
def start_from_app(app):
|
||
"""由 create_app 调用。遵守 reloader 与显式禁用开关。"""
|
||
if os.environ.get("WB_DISABLE_SCHEDULER") == "1":
|
||
app.logger.info("WB_DISABLE_SCHEDULER=1，调度器未启动")
|
||
return None
|
||
if app.debug and os.environ.get("WERKZEUG_RUN_MAIN") != "true":
|
||
return None # reloader 的父进程不启动，避免跑两份
|
||
sch = get_scheduler()
|
||
sch.start()
|
||
return sch
|