收窄熔断判断范围，修改熔断逻辑 (#2145)

zhenxun/services/db_context/config.py

@@ -27,9 +27,9 @@ MYSQL_CONFIG = {
 
 SQLITE_CONFIG = {
     "journal_mode": "WAL",  # 提高并发写入性能
-    # SQLite 的底层锁等待应接近业务超时，避免上层放弃后 worker 仍长时间占线。
+    # SQLite 底层锁等待不应超过上层 DB_TIMEOUT_SECONDS=3.0。
     # Windows bind mount / Docker 场景下 SQLite 不适合高写并发。
-    "busy_timeout": 5000,
+    "busy_timeout": 3000,
     "foreign_keys": "ON",
 }
 

zhenxun/services/db_context/utils.py

@@ -15,6 +15,12 @@ _SQLITE_STALL_UNTIL = 0.0
 _SQLITE_STALL_REASON = ""
 _DB_UNHEALTHY_TIMEOUT_SECONDS = 30.0
 _SQLITE_STALL_TIMEOUT_SECONDS = 60.0
+_SQLITE_LOCK_PATTERNS = (
+    "database is locked",
+    "database is busy",
+    "database table is locked",
+    "database table is busy",
+)
 
 
 def _is_sqlite_connection() -> bool:
@@ -46,6 +52,24 @@ def sqlite_stall_reason() -> str:
     return _SQLITE_STALL_REASON
 
 
+def _is_sqlite_lock_error(exc: BaseException) -> bool:
+    if not _is_sqlite_connection():
+        return False
+    message = str(exc).casefold()
+    return any(pattern in message for pattern in _SQLITE_LOCK_PATTERNS)
+
+
+def _mark_sqlite_lock_unhealthy(exc: BaseException, operation: str | None) -> None:
+    reason = f"{operation or 'database_operation'} sqlite lock: {exc}"
+    _mark_sqlite_stall(reason, _SQLITE_STALL_TIMEOUT_SECONDS)
+    signal_db_unhealthy(_SQLITE_STALL_TIMEOUT_SECONDS, reason=reason)
+    logger.warning(
+        "SQLite 数据库锁等待失败，已暂停低优先级数据库任务",
+        LOG_COMMAND,
+        e=exc if isinstance(exc, Exception) else None,
+    )
+
+
 async def with_db_timeout(
     coro,
     timeout: float = DB_TIMEOUT_SECONDS,
@@ -78,3 +102,7 @@ async def with_db_timeout(
                 LOG_COMMAND,
             )
         raise
+    except Exception as exc:
+        if _is_sqlite_lock_error(exc):
+            _mark_sqlite_lock_unhealthy(exc, operation)
+        raise

zhenxun/services/low_priority_writer.py

@@ -8,7 +8,7 @@ import time
 from typing import Any
 
 from zhenxun.services.log import logger
-from zhenxun.services.message_load import should_pause_tasks, signal_db_unhealthy
+from zhenxun.services.message_load import should_pause_db_tasks, signal_db_unhealthy
 from zhenxun.utils.manager.priority_manager import PriorityLifecycle
 
 LOG_COMMAND = "LowPriorityWriter"
@@ -159,7 +159,7 @@ async def _worker_loop() -> None:
             except asyncio.TimeoutError:
                 pass
             event.clear()
-        if should_pause_tasks():
+        if should_pause_db_tasks():
             continue
         async with _FLUSH_LOCK:
             for state in list(_WRITERS.values()):
@@ -186,7 +186,7 @@ async def _flush_state(
     force: bool = False,
 ) -> int:
     if not force:
-        if should_pause_tasks():
+        if should_pause_db_tasks():
             return 0
         if time.monotonic() < state.backoff_until:
             return 0

zhenxun/services/message_load.py

@@ -57,3 +57,7 @@ def db_unhealthy_reason() -> str:
 
 def should_pause_tasks() -> bool:
     return is_overloaded() or is_db_unhealthy()
+
+
+def should_pause_db_tasks() -> bool:
+    return is_db_unhealthy()