sync scripts/mofin_db.py with root + knowledge log
This commit is contained in:
知微
@@ -202,3 +202,60 @@ bash包装器启动bot会绕开systemd管理，导致：
|
||||
- 任何进程被 kill 后，下一个连接的脚本自动 checkpoint 清理 WAL，不会卡死
|
||||
- price_monitor 进程锁防止同一脚本并发
|
||||
- priority 降低减少恶性抢占
|
||||
|
||||
## 2026-07-14 10:55 — DB死锁根治：盘中脚本统一get_conn()
|
||||
|
||||
### 发现问题
|
||||
price_monitor.py的`refresh_data_prices()`已有完善的重试机制(5次+指数退避+WAL checkpoint)，但其他盘中高频脚本仍然使用raw `sqlite3.connect()`，绕过`get_conn()`的WAL模式+busy_timeout=30s设置。
|
||||
|
||||
### 修改了什么
|
||||
- `mofin_health.py` — get_db_stats()改get_conn()
|
||||
- `cron_health_monitor.py` — price_monitor新鲜度查询改get_conn()
|
||||
- `intraday_health_check.py` — 3处sqlite3.connect全部改get_conn()
|
||||
- `self_todo_executor.py` — sqlite3.connect改get_conn()
|
||||
- `mofin_db.py` — get_conn()每次新建连接时WAL checkpoint
|
||||
|
||||
### 测试结果
|
||||
并发压力测试：3x price_monitor + 3x mofin_health同时运行 → 6/6通过，WAL仅32bytes
|
||||
|
||||
### 文件
|
||||
- Modified: /home/hmo/MoFin/mofin_db.py
|
||||
- Modified: /home/hmo/MoFin/scripts/cron_health_monitor.py
|
||||
- Modified: /home/hmo/MoFin/scripts/intraday_health_check.py
|
||||
- Modified: /home/hmo/MoFin/scripts/mofin_health.py
|
||||
- Modified: /home/hmo/MoFin/scripts/self_todo_executor.py
|
||||
- Modified: /home/hmo/.hermes/profiles/position-analyst/scripts/ (同上5文件)
|
||||
- Commit: 59e94d1, pushed to origin/master
|
||||
|
||||
## 2026-07-14 10:55 — 知微XMPP Bot断线不重连修复
|
||||
|
||||
### 发现问题
|
||||
自愈执行器发现知微XMPP Bot离线。调查发现：
|
||||
1. Bot进程(3643320)由gateway产生，但无任何TCP连接（断线后slixmpp不会自动重连，因为`auto_reconnect`默认关闭）
|
||||
2. systemd的xmpp-zhiwei.service有2276次失败重启记录（PID文件冲突）
|
||||
3. Bot进程(3643320)始于04:28，presence=unavailable，但HTTP桥(5805)仍活着
|
||||
|
||||
### 根因
|
||||
slixmpp ClientXMPP的`auto_reconnect`属性默认为False。断线（connection_lost）发生后，bot进程保持运行（HTTP桥继续工作）但从不尝试重新连接XMPP。在main()中`await bot.ready.wait()`已通过，asyncio.gather继续运行，bot看起来活着但实际上无法接收/发送XMPP消息。
|
||||
|
||||
### 修改了什么
|
||||
- `/home/hmo/xmpp_agent_core.py`:
|
||||
1. AgentBot.__init__()加 `self.auto_reconnect = True`
|
||||
2. 新增 `connection_watchdog()` 看门狗协程——每15秒检查连接状态，断线超过120秒执行`os._exit(1)`让gateway重新spawn
|
||||
3. main()启动connection_watchdog作为独立任务
|
||||
|
||||
### 执行了什么
|
||||
1. Kill旧bot(3643320) → gateway自动spawn新bot(114609) → 新bot连接XMPP成功但被ejabberd重启踢下线
|
||||
2. Kill中间bot(114609) → gateway再spawn(122594) → 稳定连接
|
||||
3. Kill 122594后gateway spawn新bot(131908)加载修复代码，自动重连生效
|
||||
4. Stop systemd xmpp-zhiwei.service（2276次失败重启，不再需要，由gateway管理bot）
|
||||
5. 清理PID文件 /tmp/xmpp_zhiwei_bot.pid
|
||||
|
||||
### 效果预期
|
||||
- 断线后slixmpp自动重连（最快5s，指数退避）
|
||||
- 看门狗兜底：120秒未恢复则进程退出→gateway respawn
|
||||
- systemd不再干扰（已stop，gateway管理bot生命周期）
|
||||
|
||||
### 文件
|
||||
- Modified: /home/hmo/xmpp_agent_core.py
|
||||
- Stopped: systemctl stop xmpp-zhiwei.service (2276失败重启)
|
||||
|
||||
+21
-7
@@ -21,7 +21,7 @@ from datetime import datetime
|
||||
from pathlib import Path
|
||||
from typing import Optional, Callable
|
||||
|
||||
DATA_DIR = Path(__file__).parent.parent / "data"
|
||||
DATA_DIR = Path(__file__).parent / "data"
|
||||
DB_PATH = DATA_DIR / "mofin.db"
|
||||
|
||||
# ═══════════════════════════════════════════════════════════
|
||||
@@ -1088,8 +1088,19 @@ def write_holding_strategy(conn, code: str, name: str, data: dict) -> tuple[bool
|
||||
quality_issues_j = _json.dumps(data.get('quality_issues', {}), ensure_ascii=False) if isinstance(data.get('quality_issues'), dict) else data.get('quality_issues_json', '')
|
||||
signal_factors_j = _json.dumps(data.get('signal_factors', []), ensure_ascii=False) if isinstance(data.get('signal_factors'), list) else data.get('signal_factors_json', '')
|
||||
|
||||
# 在DELETE前保留现有的full_analysis和reassessed_at（防止被regenerate_all等清空）
|
||||
_existing_fa = data.get('full_analysis', '')
|
||||
_existing_ra = data.get('reassessed_at', '')
|
||||
if not _existing_fa:
|
||||
try:
|
||||
_old = conn.execute("SELECT full_analysis, reassessed_at FROM holding_strategies WHERE code=? ORDER BY id DESC LIMIT 1", (code,)).fetchone()
|
||||
if _old:
|
||||
if _old[0]: _existing_fa = _old[0]
|
||||
if _old[1]: _existing_ra = _old[1]
|
||||
except:
|
||||
pass
|
||||
|
||||
# DELETE + INSERT
|
||||
conn.execute("PRAGMA foreign_keys=OFF") # 临时禁用FK（自选股可能不在stocks表）
|
||||
conn.execute("DELETE FROM holding_strategies WHERE code=?", (code,))
|
||||
conn.execute("""
|
||||
INSERT INTO holding_strategies
|
||||
@@ -1100,10 +1111,11 @@ def write_holding_strategy(conn, code: str, name: str, data: dict) -> tuple[bool
|
||||
source, reason, updated_at,
|
||||
avg_price, decision_timestamp, note, quality_check,
|
||||
quality_checked_at, quality_issues_json, position_advice,
|
||||
signal_factors_json, time_horizon, decision_type, full_analysis)
|
||||
signal_factors_json, time_horizon, decision_type,
|
||||
full_analysis, reassessed_at)
|
||||
VALUES (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,
|
||||
datetime('now','localtime'),
|
||||
?,?,?,?,?,?,?,?,?,?,?)
|
||||
?,?,?,?,?,?,?,?,?,?,?,?)
|
||||
""", (
|
||||
code, name,
|
||||
data.get('version', 1), data.get('price'), data.get('cost'),
|
||||
@@ -1119,14 +1131,16 @@ def write_holding_strategy(conn, code: str, name: str, data: dict) -> tuple[bool
|
||||
data.get('avg_price', 0),
|
||||
data.get('timestamp') or data.get('created_at', ''),
|
||||
data.get('note', ''),
|
||||
data.get('quality_check', 'pending'),
|
||||
data.get('quality_check', ''),
|
||||
data.get('quality_checked_at', ''),
|
||||
quality_issues_j,
|
||||
data.get('position_advice', ''),
|
||||
signal_factors_j,
|
||||
data.get('time_horizon', ''),
|
||||
data.get('decision_type', data.get('strategy_type', 'holding')),
|
||||
data.get('full_analysis', ''),
|
||||
data.get('type', data.get('strategy_type', 'holding')),
|
||||
# 保留full_analysis和reassessed_at
|
||||
_existing_fa,
|
||||
_existing_ra,
|
||||
))
|
||||
conn.commit()
|
||||
return True, f"策略 {code} 已写入"
|
||||
|
||||
Reference in New Issue
Block a user